CN102179375A - Nondestructive detecting and screening method based on near-infrared for crop single-grain components - Google Patents

Abstract

The invention relates to a nondestructive detecting and screening method based on near-infrared for crop single-grain components, and the method comprises the following steps: acquiring a database for single-grain components of different materials, by utilizing a known ration detection process for the chemical components of crop single-grain; collecting near-infrared spectrum information of the known single-grain component materials; establishing a component identifying model corresponding to the single-grain component and the near-infrared spectrum information; and in the detection process, detecting the spectrum of a to-be-detected sample firstly, and then comparing with the identifying model and analyzing so as to acquire the component of the to-be-detected sample, and lastly automatically screening. The high-throughput nondestructive detection on chemical matters and chemical pollutants is realized on the basis of single-grain level, wherein the chemical matters contain starch, protein, fat, and the like. The method can be used for nondestructively selecting mutant and genetic segregation groups and detecting pollutants, thereby supplying a new method for ensuring the safety of crop heredity breading and the safety of agricultural products.

Description

A non-destructive detection and screening method for single-grain components of crops based on near-infrared

technical field

The invention relates to the technical field of non-destructive detection of crops, in particular to a method for non-destructive detection and screening of single grain components of crops based on near infrared.

technical background

Nondestructive Determination Technologies (NDT for short) is an emerging comprehensive application discipline, which utilizes the heat, sound, and light caused by the abnormal internal structure of the sample or the existence of defects on the premise of not destroying or damaging the object to be tested. , electric, magnetic and other changes to detect its internal and surface defects, and make judgments and evaluations on the type, nature, quantity, shape, position, size, distribution and changes of defects. According to the different non-destructive testing principles, the testing methods can be roughly divided into optical characteristic analysis method, acoustic characteristic analysis method, machine vision technology detection method, electrical characteristic analysis method, nuclear magnetic resonance detection technology and X-ray detection technology, etc.

Infrared spectroscopy and Raman spectroscopy in the optical property analysis method can conduct non-destructive analysis of samples, and have the characteristics of non-contact, non-destructive, high detection sensitivity, short time, small amount of sample required and no need for sample preparation. During the analysis process, it will not cause chemical, mechanical, photochemical and thermal decomposition of the sample, which is one of the research hotspots in the field of analytical science.

In recent years, near-infrared spectroscopy has been widely used in the non-destructive testing of agricultural products, especially in the quality analysis of crops and pesticide residues.

In foreign countries, as early as the early 1960s, Norris et al., the instrument laboratory of the US Department of Agriculture, first used near-infrared spectroscopy to measure the content of moisture, protein, fat, etc. in grains, and devoted themselves to the application of near-infrared spectroscopy in the quality analysis of agricultural and animal husbandry products. Research. So far, many experimental schemes and calculation methods of near-infrared spectroscopy have become the standard methods of AOCA (Association of Official Analytical Chemists). The American Association of Cereal Chemistry approved the near-infrared method for the determination of wheat protein in October 1982, and the International Association of Cereal Science and Technology stipulated a detailed procedure for the determination of protein and moisture content in wheat and flour by near-infrared. The practical rules of qualitative and quantitative analysis standards based on chemometrics have been published by ASTM (American Society for Testing and Materials) in 1995 and 1996. In terms of pesticide residues, Saranwong et al. (2005, 2007) used the dry sample extraction (DESIR) sample pretreatment technology to conduct a near-infrared spectrum detection study on the fungicide fungicide. The concentration of the pesticide samples ranged from 2 to 90 μg*mL-1 (ppm), and the concentration interval was 2 μg*mL ^-1 . When 2mL of the sample solution is added to the filter paper in the polystyrene petri dish, and then the solution in the petri dish is dried and the diffuse reflectance spectrum of the filter paper is directly measured, the effect of the established near-infrared spectrum analysis model is the best, and the calibration model The SEP is 6158 μg*mL ^-1 .

In China, the application research of near-infrared spectroscopy technology in crop quality analysis is also very active. Yan Yanlu et al. (1990) applied the Fourier transform near-infrared diffuse reflectance spectroscopic analysis method to measure 17 components such as protein, amino acid and fat in millet, corn, wheat and other crops, and achieved good results. Li Daqun et al. (1990) used near-infrared diffuse reflectance spectroscopy to measure the protein content of soybean and wheat. They used 35 soybean varieties and 77 wheat varieties as calibration samples to establish a prediction model. The difference between the near-infrared measured value and the actual value The correlation coefficients were 0.967 and 0.984, and the standard deviations were 0.826 and 0.348, respectively. Peng Yukui et al. (1997) compared and determined the grain quality components of 124 wheat varieties with near-infrared methods, and showed that the moisture, crude protein, crude fiber, and lysine content of wheat samples measured by near-infrared spectral analysis technology were different from conventional methods. There is a high degree of correlation between the results. Wang Cheng (2000) used Fourier transform near-infrared diffuse reflectance spectroscopy to measure the crude protein content of barley grains, and established a prediction mathematical model with 40 barley samples. The correlation coefficient between the predicted value and the measured value was 0.989. The correlation coefficient between the measured value and the measured value is 0.969, which proves that the established model has good prediction accuracy. The determination of amylose and amylopectin content in rice seeds, the determination of fat components in soybeans and rapeseed, and the determination of vitamins and sugars in fruits and vegetables have all successfully applied near-infrared spectroscopy analysis technology. In terms of pesticide residues, Xuemei et al. (2007) also used the sample pretreatment method of enrichment and purification of silica gel to absorb low-concentration ethyl carbamate into silica gel, and measured its near-infrared diffuse reflectance spectrum. The PLS model established by the band, the cross-validation error of 20 samples within the concentration of 0100～1100mg*L ^-1 is 011152mg*L ^-1 . Shen Fei et al. (2009) used near-infrared spectroscopy for the quantitative detection of trace pesticide phoxim directly.

The progress in single-seed detection of crop grain components is mainly to use near-infrared spectroscopy to detect the quality components of complete single seeds without destroying samples and fast measurement speed. Delwiche (1998) studied the feasibility of non-destructive determination of wheat single seed protein content by near-infrared method. Velasco et al. (1999, 2002) used near-infrared reflection analysis technology to non-destructively measure the fatty acid components of sunflower single intact seeds, the oil content, fat components and protein content of rapeseed seeds, and considered that near-infrared non-destructive analysis Reliable results can be obtained. Zhang Huahui et al. (1998) used near-Fourier transform infrared spectroscopy to measure the oil content of intact single corn seeds without damage, and obtained positive results.

At present, the technology for non-destructive high-throughput detection at the single-grain level includes a photoelectric color sorting device based on the photoelectric principle to sort the color difference of the single-grain surface, and it has been successfully applied to rice selection. The photoelectric color sorter uses the optical and colorimetric properties of the material to non-destructively detect defective products and sundries with abnormal colors or surface defects from a large number of bulk sample materials, and automatically sort and reject them. , which comprehensively applies new technologies such as electronics and biology, and is a typical high-tech equipment integrating light, machinery and electricity. Since the photoelectric color sorter sorts by color, it can greatly improve the quality of materials, so the unique role of adapting to the commodity market will be very obvious.

However, so far, there has been no report of a high-throughput non-destructive single-grain detection and sorting method that combines near-infrared detection of chemical composition of crop grains with automatic continuous sorting.

Contents of the invention

The invention provides a non-destructive detection and screening method for single-seed crop components based on near-infrared, which can sort individuals with different chemical components at the level of single seeds. High-throughput non-destructive detection of starch, protein, fat and other chemical substances and chemical pollutants can be realized at the single-grain level. This invention can perform non-destructive selection and pollutant detection on mutants and genetically isolated populations, and provide a basis for crop genetic breeding and agricultural products. Security offers new approaches.

The principle of near-infrared non-destructive testing method:

Near-infrared spectroscopy mainly obtains information on the structure, composition, and state of molecules through frequency-doubling and combined-frequency absorption spectra of organic molecules, and from near-infrared reflection spectroscopy, the density, particle size, polymerization degree of polymers, and fiber density of samples can also be obtained. The physical state information of matter such as diameter. Because the absorption in the near-infrared spectral region is mainly the frequency multiplication and combined frequency absorption of the molecular or atomic vibration fundamental frequency above 2000cm ^-1 , the near-infrared spectrum of organic matter mainly includes the frequency multiplication and combination of hydrogen-containing groups such as CH, NH, and OH. frequency absorption band. The absorption frequency of these hydrogen-containing groups is highly characteristic, less affected by the internal and external environment of the molecule, and the spectral characteristics of the sample in the near-infrared spectral region are very stable. This essential feature of near-infrared spectroscopy lays the foundation for the rapid identification of the chemical composition of crop seeds. The present invention applies near-infrared spectroscopy technology to the non-destructive detection of the chemical composition of single grains, performs correlation analysis on crop samples of known chemical composition and spectral data, establishes a chemical composition identification model for single grains and groups, and utilizes a large number of bulk Based on the principle of continuous and rapid high-throughput selection technology in sample materials, a near-infrared high-throughput non-destructive selection platform device is established to realize high-throughput non-destructive selection of mutants and genetically isolated populations.

To achieve the above object, the present invention adopts the following technical solutions:

A non-destructive detection and screening method based on near-infrared crop single-grain components, characterized in that it specifically includes the following steps:

(1) Establishment of standard or reference sample single particle spectrum set:

First, carry out near-infrared detection on various standard crop seed samples, and convert the crop seed spectral images into sample spectral basic data; measure the spectra of various standard crop seed samples; The measurement needs to be repeated, and the average spectrum is approximated as the standard spectrum of the sample;

(2) Establishment of single grain composition database of different materials:

Quantitative detection of chemical components of various standard crop seed samples, so as to establish a single-grain component database of various standard crop seed samples;

(3) Establishment of identification model for single particle composition of different materials:

Use the corresponding statistical analysis software to analyze the correlation between the single grain composition data and spectral information of various standard crop seed samples, and establish the identification model of single grain composition and spectral information of various standard crop seed samples; determine the standard crop seed samples Each chemical composition contained in a single seed and its spectral threshold value, which is the standard for identifying whether the chemical composition is contained in the new crop seed to be identified;

(4) Analysis of single grain components of crop seeds to be identified:

Collect the near-infrared spectral information of the single grain of the crop seeds to be identified that are being sorted on the conveyor belt, and then compare and analyze the spectral information corresponding to the components of the standard crop seeds in the identification model. whether it contains a certain chemical ingredient;

(5) Automatic sorting of samples:

If the near-infrared spectrum of the crop seeds to be identified is compared with the identification model and meets the set expectations, the crop seed samples to be identified will directly enter the good product collection area under the conveyance of the conveyor belt; otherwise, when the crop seed samples to be identified arrive at the sorting area, spray The valve ejects high-speed airflow to blow it into the defective product collection area, so as to realize the automatic screening of samples;

The crop seeds are rice, wheat, corn and the like.

The spectral information has undergone the following corrections and preprocessing:

After obtaining the spectral information, perform spectral correction to normalize the spectral graph, offset background interference and improve the quality of the spectrum, using one of smoothing, centering, derivation, normalization, multiple scattering correction, SNV, Reduce, Noise, Any two or any three kinds of spectral preprocessing, which correction method to use should be selected according to the quality of the spectrum and the interference situation. Preprocessing can also amplify the original hidden signal difference, improve the resolution of the spectrum, and make the variety The identification is more intuitive and reliable;

Described comparative analysis is the qualitative discriminant analysis of identifying crop seed varieties in the near-infrared spectrum, and more relies on several peak groups or frequency segments or even the whole spectrum to carry out qualitative discrimination, including deviation weight method, Kruskal-Wallisting test, principal component analysis, Partial least squares method, DPLS, SIMCA, LLM, Fisher discriminant, KNN, wavelet analysis or ANN feature screening method to extract spectral features to improve the reliability of analysis and identification results.

Described identification model refers to: for the crop seed sample of single grain, determine that unknown sample belongs to a certain kind, adopt pattern recognition to carry out identification, the pattern recognition method of identification crop seed variety uses Fisher discriminant, Bayes discriminant, stepwise discriminant, linear learning machine, KNN, SIMCA, DPLS, cluster analysis, least squares regression, Euclidean distance or neural network; for discriminant analysis by pattern recognition, it is necessary to divide the spectra of standard crop seed samples with known different chemical compositions into a learning set and a test set Two parts, the division is based on the fact that the categories of the learning set and the test set should be the same and have a wide range of representativeness; then assign initial values to the crop seed samples with different chemical components according to prior knowledge to establish different component identification models, and then use The test set is used to evaluate the performance of the model.

The beneficial effect that the present invention has compared with background technology is:

1) Using spectral technology to identify chemical components and pesticide residues of crop seeds, the analysis speed is greatly accelerated. The measurement process of the spectrum can generally be completed within 30 seconds;

2) No chemical reagents are used, which reduces the detection cost and does not pollute the environment;

3) Compared with the chemical method, the system error and human error are greatly reduced, and the measurement accuracy is improved;

4) It can handle a large number of samples and single sample analysis, saving time, and the real-time detection technology can track and detect mutants and genetic groups very well;

5) The analysis sample can be identified non-destructively, and the identified crops can still be used for planting and production.

Description of drawings

Figure 1 is a typical spectrum curve of a single crop seed.

Fig. 2 is a flow block diagram of the method of the present invention.

Detailed ways

First, the peripherals are powered on, the computer is controlled to start up, the system control unit is started, and the near-infrared nondestructive testing system, the transport unit and the sorting unit are initialized. After the initialization is completed, the seed delivery unit of the system arranges the seeds to be detected on the conveyor belt neatly, and the conveyor belt moves the first seed to the near-infrared detection unit to obtain its corresponding near-infrared spectrum. Then compare the map with the standard model library established in the early stage to judge whether it meets the requirements. After that, the conveyor belt continues to move forward, passing the next seed to the near-infrared detection unit, while the already detected seeds are sent to the sorting unit. If the requirements are not met, the seeds will be blown into a receiving container A by the high-speed airflow ejected by a special high-speed micro spray valve, otherwise, the seeds will be directly sent to another receiving container B along with the movement of the conveyor belt. The system runs cyclically, so as to realize the automatic non-destructive detection of plant seeds. In order to improve the detection efficiency, it can be realized by increasing the number of channels.

Claims (4)

1. A non-destructive detection and screening method based on near-infrared crop single-grain components, characterized in that: it specifically comprises the following steps: (1) Establishment of standard or reference sample single particle spectrum set: First, perform near-infrared detection on various standard crop seed samples, and convert the spectral image into the basic data of the sample spectrum; measure the spectra of various standard crop seed samples; generally the same sample needs to be measured repeatedly, and samples of different batch numbers also need to be repeated For measurement, the average spectrum is approximated as the standard spectrum of the sample to establish a standard or reference sample single particle spectrum set; (2) Establishment of single grain composition database of different materials: Quantitative detection of chemical components of various standard crop seed samples, so as to establish a single-grain component database of various standard crop seed samples; (3) Establishment of identification model for single particle composition of different materials: Use the corresponding statistical analysis software to analyze the correlation between the single grain composition data and spectral information of various standard crop seed samples, and establish the identification model of single grain composition and spectral information of various standard crop seed samples; determine the standard crop seed samples Each chemical composition contained in a single seed and its spectral threshold value, which is the standard for identifying whether the chemical composition is contained in the new crop seed to be identified; (4) Analysis of single grain components of crop seeds to be identified: Collect the near-infrared spectral information of the single grain of the crop seeds to be identified that are being sorted on the conveyor belt, and then compare and analyze the spectral information corresponding to the components of the standard crop seeds in the identification model. Whether it contains a certain chemical composition and its size; (5) Automatic sorting of samples: If the near-infrared spectrum of the crop seeds to be identified is compared with the identification model and meets the set expectations, the crop seed samples to be identified will directly enter the good product collection area under the conveyance of the conveyor belt; otherwise, when the crop seed samples to be identified arrive at the sorting area, spray The valve ejects high-speed airflow to blow it into the defective product collection area, so as to realize the automatic screening of samples; The crop seeds are rice, wheat, corn and the like. 2. The non-destructive detection and screening method based on near-infrared crop single-grain components according to claim 1, characterized in that: The spectral information has undergone the following corrections and preprocessing: After obtaining the spectral information, perform spectral correction to normalize the spectral graph, offset background interference and improve the quality of the spectrum, using one of smoothing, centering, derivation, normalization, multiple scattering correction, SNV, Reduce, Noise, Any two or any three kinds of spectral preprocessing, which correction method to use should be selected according to the quality of the spectrum and the interference situation. Preprocessing can also amplify the original hidden signal difference, improve the resolution of the spectrum, and make the variety The identification is more intuitive and reliable; 3. The non-destructive detection and screening method based on near-infrared crop single-grain components according to claim 1, characterized in that: Described comparative analysis is the qualitative discriminant analysis of identifying crop seed varieties in the near-infrared spectrum, and more relies on several peak groups or frequency segments or even the whole spectrum to carry out qualitative discrimination, including deviation weight method, Kruskal-Wallisting test, principal component analysis, Partial least squares method, DPLS, SIMCA, LLM, Fisher discriminant, KNN, wavelet analysis or ANN feature screening method to extract spectral features to improve the reliability of analysis and identification results. 4. The non-destructive detection and screening method based on near-infrared crop single-grain components according to claim 1, characterized in that: Described identification model refers to: for the crop seed sample of single grain, determine that unknown sample belongs to a certain kind, adopt pattern recognition to carry out identification, the pattern recognition method of identification crop seed variety uses Fisher discriminant, Bayes discriminant, stepwise discriminant, linear learning machine, KNN, SIMCA, DPLS, cluster analysis, least squares regression, Euclidean distance or neural network; for discriminant analysis by pattern recognition, it is necessary to divide the spectra of standard crop seed samples with known different chemical compositions into a learning set and a test set Two parts, the division is based on the fact that the categories of the learning set and the test set should be the same and have a wide range of representativeness; then assign initial values to the crop seed samples with different chemical components according to prior knowledge to establish different component identification models, and then use The test set is used to evaluate the performance of the model.

Images