CN114527082A

CN114527082A - Sweet corn seed germination prediction method and device

Info

Publication number: CN114527082A
Application number: CN202210067241.2A
Authority: CN
Inventors: 崔华威; 邴阳; 张晓迪; 王子麟; 李龙威; 苗爱敏
Original assignee: Zhongkai University of Agriculture and Engineering
Current assignee: Zhongkai University of Agriculture and Engineering
Priority date: 2022-01-20
Filing date: 2022-01-20
Publication date: 2022-05-24

Abstract

The embodiment of the invention relates to the technical field of crop seed growth prediction, and discloses a sweet corn seed germination prediction method, which comprises the following steps: acquiring hyperspectral test data of seeds to be tested; preprocessing the hyperspectral test data to obtain hyperspectral characteristic data; performing first order difference on the hyperspectral characteristic data by adopting a first order difference model to obtain key spectral characteristics related to seed vigor; and calculating according to the key wave band characteristics and a regression analysis algorithm to determine the vitality state of the seeds. The invention also discloses a sweet corn seed germination prediction model. In the embodiment of the invention, relevance is established by the hyperspectral data of the seeds and the root length of the seedlings after germination, key spectral information of hyperspectrum is obtained by differential processing aiming at hyperspectral characteristics, and then regression analysis is adopted to predict the root length of the seedlings; the method can realize rapid, nondestructive and high-precision seed viability detection.

Description

Sweet corn seed germination prediction method and device

Technical Field

The invention relates to the technical field of crop seed growth prediction, in particular to a sweet corn seed germination prediction method and device.

Background

Sweet corn (Zea mays l.saccharata) is currently a vegetable crop of high nutritional and edible value, as it is rich in sugars, various amino acids, vitamins, minerals and dietary fibres. Sweet corn is various in variety and is more popular with consumers all over the world than common corn. According to introduction, the planting area of sweet corn in China is gradually enlarged in recent years. In 2018, the planting area of the sweet corn in China exceeds 3000 square kilometers, and the planting area accounts for 25% of the total world production. With the continuous improvement of the requirements of people on safe production and variety reliability, high-quality seeds become a great importance for the development of the planting industry. However, the vegetation (soil moisture, temperature, nutrition, pests), harvest (mechanical damage, maturity) and post harvest (drying and storage of seeds) environmental conditions of seed planting are difficult to control. These factors may lead to severe loss of seed growth and development and developmental retardation. The determination of the seed vitality is the key point of modern seed science and is also the premise of high yield. The seed vigor is an important index for measuring the comprehensive germination rate, emergence rate, seedling growth potential and plant stress resistance of seeds. Therefore, the seed vigor must be known before sowing to ensure a higher germination rate and economic benefits. Therefore, the method for detecting the seed vigor is established quickly, nondestructively and accurately, and has important biological and economic significance for ensuring the seed quality, optimizing crop production facilities and improving the crop yield.

Disclosure of Invention

Aiming at the defects, the embodiment of the invention discloses a sweet corn seed germination prediction method, which adopts hyperspectral data to predict the seedling root length, can realize better seed germination prediction and can realize better nondestructive detection.

The embodiment of the invention discloses a sweet corn seed germination prediction method in the first aspect, which comprises the following steps:

a sweet corn seed germination prediction method comprises the following steps:

acquiring hyperspectral test data of seeds to be tested;

preprocessing the hyperspectral test data to obtain seed hyperspectral characteristic data;

performing characteristic acquisition on the hyperspectral data by adopting a spectrum first-order differential model to determine corresponding key waveband characteristics;

and calculating according to the key wave band characteristics and a regression analysis algorithm to determine the vitality state of the seeds.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the preprocessing the hyperspectral test data to obtain the hyperspectral characteristic data includes:

extracting an interested region of the hyperspectral test data by adopting an elliptical segmentation mode to obtain interested feature data;

performing black-and-white correction on the interested characteristic data by adopting a black-and-white correction formula to obtain preprocessed hyperspectral characteristic data, wherein the black-and-white correction formula is as follows:

wherein I is hyperspectral characteristic data, I_rawFor collecting seed hyperspectral data, I_blackFor dark-light correction data obtained when the scanning lens is blocked, I_whiteWhite light correction data was obtained by scanning a calibrated white correction plate with a reflectance of 99.99%.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the hyperspectral measured data is hyperspectral data within a spectral range of 400nm to 1000 nm; the hyperspectral test data is one of hyperspectral data of the embryo surface of the seed or hyperspectral data of the embryo breast surface of the seed.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the performing feature extraction on the hyperspectral characteristic data by using the first-order difference information acquisition model to determine corresponding key spectral features includes:

performing characteristic waveband division on the hyperspectral characteristic data according to a preset number of spectral bands;

constructing original spectral data and corresponding first-order difference characteristics thereof to determine the incidence relation between the spectral characteristics and the seedling root length; the correlation coefficient model is:

assume that the original data sequence is y_i(1 ≦ i ≦ n), the variable y is defined only on non-negative integer values; when the variable i sequentially takes non-negative integers, and when i is changed from k to k +1, the change amount of the dependent variable is as follows:

△y_k＝y_k+1-y_k (2)

△y_kreferred to as the first difference of the function at point k;

spectral feature information is determined based on the spectral data and its first order difference information.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the number of variables of the raw spectral band data is preset to be 220.

As an alternative implementation manner, in the first aspect of the embodiment of the present invention, the calculating according to the key band features and a regression analysis algorithm to determine the seed vigor state includes:

and performing regression analysis on the key characteristic wave band by adopting a principal component regression, partial least square regression and support vector regression algorithm to determine the seed vigor state, wherein the seed vigor state is the seedling root length.

As an optional implementation manner, in the first aspect of the embodiment of the present invention, the kernel function selected by the support vector regression algorithm is a gaussian radial basis function; the gaussian radial basis function is:

k(x,y)＝exp(-||x-y||²k (x, y), where σ is selected to be 50.

The second aspect of the embodiment of the invention discloses a sweet corn seed germination prediction device, which comprises:

an acquisition module: the hyperspectral testing data acquisition unit is used for acquiring hyperspectral testing data of the seeds to be tested;

a preprocessing module: the hyperspectral data processing system is used for preprocessing the hyperspectral test data to obtain hyperspectral characteristic data;

a band selection module: the system comprises a spectrum first-order difference algorithm, a spectrum characteristic data acquisition module and a spectrum characteristic data acquisition module, wherein the spectrum first-order difference algorithm is used for selecting and processing the characteristics of the high spectrum characteristic data by a spectrum to determine corresponding key spectrum characteristics;

a state determination module: and the method is used for calculating according to the key wave band characteristics and a regression analysis algorithm to determine the vitality state of the seeds.

A third aspect of an embodiment of the present invention discloses an electronic device, including: a memory storing executable program code; a processor coupled with the memory; the processor calls the executable program code stored in the memory for executing the sweet corn seed germination prediction method disclosed in the first aspect of the embodiment of the invention.

In a fourth aspect of the embodiments of the present invention, a computer-readable storage medium is disclosed, which stores a computer program, wherein the computer program enables a computer to execute the sweet corn seed germination prediction method disclosed in the first aspect of the embodiments of the present invention.

Compared with the prior art, the embodiment of the invention has the following beneficial effects:

in the embodiment of the invention, relevance is established by establishing hyperspectral data and the root length of the seedling after germination, then a first-order difference model is adopted to select key spectral information of corresponding hyperspectrum, and then regression analysis is adopted to determine the root length prediction of the seedling; the method can realize rapid, nondestructive and high-precision seed viability detection.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.

FIG. 1 is a schematic flow chart of a method for predicting germination of a sweet corn seed according to an embodiment of the present invention;

FIG. 2 is a schematic flow chart of the pre-processing steps disclosed in the embodiments of the present invention;

FIG. 3 is a schematic flow chart of band selection according to an embodiment of the present invention;

FIG. 4 is a flow chart of the construction of a predictive model according to the present disclosure;

FIG. 5 is a graph showing the relationship between the hyperspectral data and the correlation coefficient of the seedling root length disclosed in the embodiment of the invention;

FIG. 6 is a graphical representation of regression prediction results of shoot root length as disclosed in an embodiment of the present invention;

FIG. 7 is a schematic structural diagram of a sweet corn seed germination prediction apparatus provided in an embodiment of the present invention;

fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first", "second", "third", "fourth", and the like in the description and the claims of the present invention are used for distinguishing different objects, and are not used for describing a specific order. The terms "comprises," "comprising," and "having," and any variations thereof, of embodiments of the present invention are intended to cover non-exclusive inclusions, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements explicitly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Conventional methods for evaluating seed viability include immunoassay tests, polymerase chain reaction tests and germination tests. However, the chemical or planting methods described above are expensive, time consuming and destructive, often requiring many instruments. Therefore, they are not suitable for use when rapid and convenient estimation of seed vigor is required. In order to achieve accurate monitoring and quality control, reliable non-destructive inspection methods are required. Recently, some non-destructive inspection methods such as X-ray diffraction, laser speckle analysis techniques, and conductivity measurement have been proposed. Seed detection applications based on these methods are limited due to low efficiency and complex operation. Fortunately, recent research has shown that spectrum-based techniques, such as near infrared spectroscopy, nuclear magnetic resonance spectroscopy, photoacoustic spectroscopy, hyperspectral, multispectral, and fourier transform near infrared spectroscopy, have been developed and successfully applied. In particular, hyperspectral imaging (HSI) is a new technology that combines traditional images with spectroscopic techniques while recording spectral and spatial information of a study object, and this feature is very important for seed detection. Therefore, this method shows a great potential in seed vigor evaluation compared to other point spectrum techniques that cannot provide spatial information. Thus, HSI has successfully identified four different varieties of cotton seeds, green coffee beans, hybrid seeds (okra and luffa), hybrid seeds, sweet corn seeds, and waxy corn seeds. HSI technology is also widely applied to corn seeds in terms of viability estimation. Taking a short-wave infrared hyperspectral camera in the range of 1000-2500 nm as an example, the short-wave infrared hyperspectral camera is analyzed. The result shows that the combination of the visible near-infrared hyperspectral imaging technology with Multiplicative Scattering Correction (MSC), Genetic Algorithm (GA) and Partial Least Squares Regression (PLSR) is a feasible and reliable method for determining the conductivity of the corn seeds. In order to detect the vitality of the corn seeds in the storage process, related technical personnel propose a method for identifying the vitality of the seeds under 8 different aging times by using HSI, and the result shows that the feasibility and the effectiveness of evaluating the vitality and the sowing degree of the seeds by using HSI. Researchers have also developed a new technique for detecting the degree of influence of microwave heat treatment on seed viability of three different varieties (yellow and white) and comparing the influence of different spectrum pretreatment methods, corn seed characteristics and spectral ranges on seed germination rate prediction. The accuracy of identifying corn seeds (heat treated) and normal (untreated) seeds by partial least squares discriminant analysis (PLS-DA) reaches 95.6%. In order to combine the spectrum and image information of the HSI to predict the vitality of the seeds, a multi-channel data acquisition system is adopted to measure the image and the spectrum. And (4) collecting hyperspectral information of 4 activity level seeds 10h before germination, and performing activity evaluation by using a convolutional neural network. By comparing different preprocessing and pattern recognition models, the convolutional neural network model has the best comprehensive spectrum and image information recognition effect, and the prediction precision of the four activity levels is higher. Moisture content prediction which directly affects storage time and seed germination. Based on hyperspectral images of both sides (embryo side and endosperm side) of each seed of 4 maize varieties, PLSR prediction was performed. Based on the above, the embodiment of the invention discloses a sweet corn seed germination prediction method, a device, electronic equipment and a storage medium, wherein relevance is established between hyperspectral data and root length after germination, then a key waveband of a corresponding hyperspectral is selected by adopting waveband selection, and then seedling root length prediction is determined by adopting regression analysis; the method can realize rapid, nondestructive and high-precision seed vigor detection.

Example one

Referring to fig. 1, fig. 1 is a schematic flow chart of a sweet corn seed germination prediction method disclosed in an embodiment of the present invention. The execution main body of the method described in the embodiment of the present invention is an execution main body composed of software or/and hardware, and the execution main body can receive related information in a wired or/and wireless manner and can send a certain instruction. Of course, it may also have certain processing and storage functions. The execution body may control a plurality of devices, such as a remote physical server or a cloud server and related software, or may be a local host or a server and related software for performing related operations on a device installed somewhere. In some scenarios, multiple storage devices may also be controlled, which may be co-located with the device or located in a different location. As shown in fig. 1, the sweet corn seed germination prediction method comprises the following steps:

s101: acquiring hyperspectral test data of seeds to be tested;

more preferably, the hyperspectral test data are hyperspectral data within a spectral range of 400nm-1000 nm; the hyperspectral test data is one of hyperspectral data of the embryo surface of the seed or hyperspectral data of the embryo breast surface of the seed.

In specific implementation, a visible/near-infrared hyperspectral imaging system is adopted, the visible/near-infrared hyperspectral imaging system has a wave band of 386.7-1016.7nm and comprises a CCD camera with the spectral resolution of 3nm +/-0.5 nm, two 50WLED illuminating lamps and an ND conveying belt driven by a stepping motor. The system is controlled by a computer with SpVIEW software. Considering the heterogeneity of corn seeds, the embryo face and the endosperm face of each seed are respectively detected. The above is a specific hyperspectral image acquisition step, but when training is performed, the spectral data of each variety can be divided into 73 training sets and 16 test sets by using the Kennard-Stone method. And then calculating the classification accuracy of the model by using the test set. Each seed was placed on a stage in the same arrangement (8 rows × 8 columns) with the embryo side up and the endosperm side up, respectively, and then scanned in 1.2mm/s rows for 15ms to obtain 256 band spectral data.

The root length of the germinated seeds reflects the activity of the seeds. Embodiments of the invention utilize this index to monitor seed vigor. Soaking the seeds in water under the same conditions, and continuously observing the germination condition of the seeds by using a paper tower germination method. The root length of the seedling was determined 7 days after seed germination, using the longest root length as a standard. And adjusting and training the model through the specific data.

S102: preprocessing the hyperspectral test data to obtain hyperspectral characteristic data;

fig. 2 is a schematic flow chart of a preprocessing step disclosed in the embodiment of the present invention, and as shown in fig. 2, more preferably, the preprocessing the hyperspectral test data to obtain hyperspectral characteristic data includes:

s1021: extracting an interested region of the hyperspectral test data in an oval segmentation mode to obtain interested feature data;

s1022: performing black-and-white correction on the interested characteristic data by adopting a black-and-white correction formula to obtain preprocessed hyperspectral characteristic data, wherein the black-and-white correction formula is as follows:

When the hyperspectral images of the sweet corn seeds are collected, the spectrum signals obtained by the spectrometer not only contain useful information, but also can be superimposed with random errors, so that the spectrum data are easy to generate noise interference, and therefore, the middle 220 wave bands from 430.1nm to 971.5nm are selected for analysis. Second, a region of interest (ROI) was created using elliptical segmentation of the embryo and endosperm portions of the seed, and then average spectral data within the ROI was extracted using ENVI 5.1. In order to reflect the actual spectrum condition of the seeds by the collected spectrum data, white light correction and dark light correction are carried out on the collected images to eliminate noise influence, the equipment is kept to collect samples under the same parameter setting, and white light correction data I is obtained by scanning a standard white correction plate with the reflectivity of 99.99 percent_whiteAnd covering the scanning lens to obtain dark light correction data I_blackUsing two reference values for the spectral raw image (I)_raw) And (3) carrying out calibration: and calculating the corrected hyperspectral data I according to the consensus.

S103: selecting the characteristics of the hyperspectral characteristic data by adopting a spectrum first-order differential model to determine corresponding key waveband characteristics;

fig. 3 is a schematic flow chart of processing and fusing the spectral data and the first-order difference characteristic data thereof, as shown in fig. 3, the determining of the corresponding comprehensive hyperspectral characteristic data by performing characteristic fusion on the hyperspectral characteristic data by using the spectral characteristic comprehensive model includes:

s1031: performing characteristic wave band division on the hyperspectral characteristic data according to a preset number of spectral wave bands;

s1032: constructing each spectral band data and the corresponding first-order difference characteristic thereof to determine the incidence relation between the spectral characteristic and the seedling root length; the correlation coefficient model is:

△y_k＝y_k+1-y_k

△y_kreferred to as the first difference of the function at point k;

determining spectral feature information based on the spectral data and first order difference information thereof;

s1033: and carrying out numerical value fusion on the obtained difference information and the original spectrum information to obtain complete spectrum characteristics.

More preferably, the number of original spectral bands is 220.

Each extracted spectrum consists of 220 original spectral bands, and 219 bands of first order difference variables ranging from 430.1 to 971.5 nm. Since a small number of variables can reduce redundancy and computational effort, fewer bands are expected to represent a large portion of useful information. In order to obtain the characteristic wave band most relevant to the vitality of the seeds, the embodiment of the invention constructs high spectral wave band data of the seeds and first-order difference information thereof, and obtains the data characteristic with higher association degree with the seedling root length. And finally establishing a regression association relation between the key characteristic wave band of the seeds and the root length of the seedlings through a regression model.

Besides the most available method, the method can be combined with a commonly used hyperspectral data preprocessing method to preprocess data based on methods such as Smoothing-filtering (SGS), Multivariate Scatter Correction (MSC), Standard Normal Variate transformation (SNV) and the like to obtain an optimal hyperspectral data processing model.

S104: and calculating according to the comprehensive spectral characteristics and a regression analysis algorithm to determine the vitality state of the seeds.

More preferably, the calculating according to the spectral features and the regression analysis algorithm to determine the seed vigor state includes:

and performing regression analysis on the key characteristic wave band by adopting a support vector regression algorithm to determine the seed vigor state, wherein the seed vigor state is the seedling root length.

More preferably, the kernel function selected by the support vector regression algorithm is a gaussian radial basis function; the gaussian radial basis function is:

k(x,y)＝exp(-||x-y||²k (x, y), where σ is selected to be 50.

In the embodiment of the invention, a regression relationship between hyperspectral data of seeds and the root length of the germinated seedling is established by adopting Principal Component Regression (PCR), Partial Least Squares (PLS) and Support Vector Regression (SVR). High-resolution spectroscopy provides a comprehensive, complete state for the external morphology and internal structure of the seed. Through the regression model, the relation between the hyperspectral data of the seeds and the corresponding root lengths of the seedlings representing the germination characteristics of the seeds is obtained, and a theoretical basis is provided for the nondestructive testing of the seed vitality.

The training data of the corn seeds is X (X)₁,...,x₁)∈R^D×n(D is 256 high spectral wavelengths). A flow chart of model construction and germination prediction is shown in fig. 4. In the implementation, there are many regression algorithms that can be used, for example, PCR is a multiple regression analysis method, and aims to solve the problem of multiple collinearity among independent variables in data regression. In PCR, the acquired training data is compressed to a low-dimensional space by PCA, and then a linear regression model is established between the projections and the output variables to represent the data relationship between the raw data and the output variables. The PLS algorithm projects the predictive variables and the observed variables to a new space through data projection, so that the projected data represent the original data as best as possible, and meanwhile, the observed variables have strong interpretative capacity on the predictive variables. The SVR algorithm aims to set the tolerance deviation and relaxation variables of the linear function by obtaining a regression plane, and to minimize the loss and the minimizationAnd obtaining an optimization model according to the width of the isolation zone so as to obtain the optimal independent variable and hidden variable characteristic fitting. The kernel function chosen by SVR is a gaussian radial basis function, i.e., k (x, y) ═ exp (— | | x-y | | | non-conducting cells²K (x, y) where σ is selected to be 50[36 ]]. The embodiment of the invention adopts three methods to carry out data regression on the hyperspectral characteristic data of the seeds and the corresponding seedling root lengths.

In the embodiments of the present invention. Due to the fact that a large number of seed spectral data variables are collected, collinearity and redundancy exist among the spectral variables, even noise and interference exist, and therefore the time for regression modeling of the spectral data is too long. The embodiment of the invention adopts three kinds of spectral data characteristic information, namely hyperspectral raw data, hyperspectral data fused with first-order difference information of the hyperspectral data, hyperspectral first-order difference information and the like, and establishes a regression model by using spectral characteristic variables. The seedling root length is an effective index for reflecting the seed vitality. The embodiment of the invention screens out the spectral characteristic information related to the vitality of the seeds on the basis of the spectral data of the seeds and the first-order difference information of the spectral data. FIG. 5 shows the correlation coefficient between each characteristic band and its first-order difference band information and the seedling root length. As can be seen from FIG. 5, the hyperspectral data has a correlation with seedling root length. More importantly, however, the correlation coefficients of the data of different wave bands and the seedling root length are different, and the spectral data and the first-order difference and the correlation performance difference of the seedling root length can form complementation. As can be seen from fig. 5, the high-frequency band data of the original spectrum data has a strong correlation with the seedling root length, and the medium-low frequency band data of the first-order difference information has a strong correlation with the seedling root length.

When the method is specifically implemented, a model based on seed hyperspectral data and seed seedling root length is established by using methods such as SVR, PCR, PLS and the like. And the seed vigor prediction performances under different data characteristics of original spectrum data, spectrum data fused with first-order difference data, spectrum first-order difference data and the like are specifically compared, and the seed vigor prediction performances are compared with results of different spectrum preprocessing methods such as MSC, SGS, SNV and the like. The regression and prediction results of shoot root length based on different pre-treatment and regression methods are shown in table 1, where the RMSE and correlation coefficient R2 criteria are given. The larger the correlation coefficient, the lower the RMSE, indicating that the higher the prediction accuracy, the stronger the correlation between the actual value and the predicted value.

From the comparison results in table 1, the seedling root length prediction based on the hyperspectral information and the hyperspectral first-order difference information is superior to other models in both evaluation criteria. It should be noted that the model regression of the SVR method is consistent with the actual output trajectory with minimal deviation. The performance of the kernel extension model is superior to the linear model PCR and PLS, which indicates that the performance of the nonlinear regression can be improved to improve the accuracy of the result prediction considering the kernel trick, since the data relationship in the seed is usually nonlinear. Non-linear models (such as SVR) exhibit better performance than linear models based on hyperspectral methods. In the methods for extracting different spectral features, the highest data correlation is obtained in all three regression models based on the spectral data and the first-order difference data thereof. The effectiveness of the hyperspectral first-order difference data for activity-related modeling is further proved by considering the first-order difference of the spectrum in the spectrum information. As can be seen from Table 1, the optimal correlation coefficient is 0.8319, which illustrates the correlation between hyperspectral data and seed root length. The result further verifies the feasibility of the hyperspectral method in seed germination prediction.

As can be seen from Table 1, the seedling root length prediction results obtained by using SVR, PLS and PCR regression algorithm based on the fusion of spectral data and difference data thereof all show the best performance, which is greater than the corresponding model prediction performance based on the original spectral data and the first-order difference data. In the PLS algorithm, the prediction accuracy obtained based on the raw spectral information, the spectral difference information and the fusion features thereof is similar, which is probably because the PLS algorithm is mainly based on the correlation analysis of data, and the correlation of the variables of the raw spectral information is similar to that of the difference information thereof, so the obtained results are also similar. Because of complementarity of the hyperspectrum and the data correlation between the first order difference information and the root length of the seedling, this is also shown in fig. 5. Namely, the information fusion of the two can reflect the spectrum association relation of the seedling root length.

Table 1 shows that the regression performance of the spectral data obtained by the conventional spectral data preprocessing methods such as MSC, SGS, SNV, etc. is poor. The reason is that the common preprocessing method can distort the spectral data, destroy the spatial distribution characteristics of the original spectral data, cause the correlation damage of the spectral value of the seed and the germination characteristics of the seed, affect the prediction performance of the seed vitality characteristics, and have large prediction error. As can be seen from Table 1, the hyperspectral data of the corn seeds have strong correlation with the root length characteristics of the seedlings after the seeds germinate. The effectiveness and feasibility of the hyperspectral technology in the aspects of seed germination characteristic prediction, high-quality seed identification, fine seed breeding and the like are verified. Meanwhile, the effectiveness of the germination characteristic prediction based on the spectral first-order difference information is verified through data correlation analysis.

TABLE 1

In order to study the prediction performance of the three regression methods based on the first-order difference information more clearly, fig. 6 shows a graph of the seedling root length prediction results based on the PCR, SVR and PLS models. It can be noted that all methods can capture the data change of the root length of the seedling because the predicted data is consistent with the actual value of most samples. Particularly, the track of the predicted value and the track of the actual value of the SVR model are well matched, and the deviation is minimum.

From the regression results of the sweet corn seeds, the hyperspectral data of the seeds have strong correlation with the root length after germination. The effectiveness of the hyperspectral data in the aspects of seed vitality prediction, high-quality seed identification, fine seed breeding and the like can be determined. Meanwhile, the effectiveness of the band selection is verified through data correlation. The result shows that the seedling root length prediction model obtained by the SVR model has the best effect. SVR is a kernel-based nonlinear regression method, which shows that the hyperspectral data distribution of seeds and the root length of seedlings are in a nonlinear relationship. Therefore, a kernel function-based nonlinear regression model is the most suitable model for corn seed vigor detection. It can reach better prediction effect.

The embodiment of the invention establishes a corn seed vigor prediction model. Model performance was compared for different pre-processing and spectral feature selection. Most of the existing methods are based on qualitative analysis and classification of corn seed vitality, such as microwave heat treatment and untreated seed separation, sowing degree evaluation, path and non-activity judgment and aging level prediction. Quantitative analysis is mainly focused on the prediction of components related to seed vigor. At present, the quantitative analysis and research on the vitality of corn seeds at home and abroad is less, in particular to sweet corn seeds. The active starch has insufficient accumulation, high sugar content, low seed activity, low emergence rate, weak growth potential in seedling stage and easy infection by pathogenic bacteria. The vitality detection method mainly focuses on common corn seeds with consistent characters and is high in precision. However, the current methods are not suitable for sweet corn because of the large differences between the dried varieties. In current seed vigor assays, field emergence rates are primarily considered, while the effectiveness of seed germination (e.g., root length and seedling root length) is ignored. The seedling emergence of the seeds is an important basis for ensuring the later growth and development of crops. In order to ensure the systematicness and integrity of vitality evaluation, the evaluation on the vitality and seedling emergence capability of the sweet corn seeds is particularly important.

The feature extraction of the hyperspectral data is particularly important for the vitality prediction and variety classification of the sweet corn. Although various spectral feature processing and evaluation methods are proposed, existing hyperspectral preprocessing methods such as Genetic Algorithm (GA) and continuous projection algorithm (SPA) are based on global data distribution, mainly focus on spatial distribution features of spectral data, and ignore the associated distribution structure of spectral data variables, such as the association between bands and spectral data and the association between spectral data in different bands. Based on the method, the applicant provides a method for realizing sample spectral information expression based on first-order difference information of spectral data, and the method is applied to seed vigor phenotype prediction research. Three different regression models are constructed based on different spectral data distributions by utilizing the characteristic correlation between the hyperspectral data and the seed vigor phenotype and are used for predicting and researching the corn seed vigor. In order to evaluate the accuracy of the model, the root length of the seed seedling is determined by a germination test as a reference. The prediction accuracy and superiority of different spectral data acquisition methods and regression algorithms are compared, and the effectiveness and feasibility of the nondestructive testing of the seed vigor are determined.

The sweet corn seed vigor prediction method based on the PCR, SVR and PLS models is used for predicting sweet corn seed vigor by utilizing hyperspectral data. The main conclusions are as follows.

(1) Based on regression models such as PCR, SVR and PLS, a regression incidence relation model of the corn seed root length and the spectrum data is established by utilizing the spectrum data and the first-order difference of the spectrum data, and finally quantitative prediction of the seed root length variable data is realized. The main conclusions are as follows:

(1) the accuracy of predicting the root length of the seed germinating seedlings by algorithms such as PCR, SVR and PLS is improved by fusing the corn seed spectral data with the first-order difference information, and more accurate prediction results can be obtained by fusing the first-order difference spectral data with data after pretreatment such as MSC, SGS and SNV and curve fitting model data.

(2) As seen from the results of the prediction models, the corresponding prediction results of different regression models are different, the accuracy of the SVR model is higher than that of the PCR model and the PLS model, the determination coefficient of the SVR model reaches 0.8319 at most, and the SVR model obtains more satisfactory results than the results of the PCR model and the PLS model (R2 is 0.8023 and 0.725 at most).

(3) According to the prediction result of the established regression model on the seedling root length characteristics, the actual seedling root length has larger correlation with the corresponding spectral data, the highest correlation is 0.8319, and the feasibility and the effectiveness of predicting the seed germination characteristics based on the spectral technology are verified. In the embodiment of the invention, relevance is established by establishing hyperspectral data and the root length after germination, then a key waveband of a corresponding hyperspectral is selected by adopting waveband selection, and then the root length prediction of the seedling is determined by adopting regression analysis; the method can realize rapid, nondestructive and high-precision seed viability detection.

Example two

Referring to fig. 7, fig. 7 is a schematic structural diagram of a sweet corn seed germination prediction device according to an embodiment of the present invention. As shown in fig. 7, the sweet corn seed germination prediction apparatus may include:

the acquisition module 21: the hyperspectral testing data acquisition unit is used for acquiring hyperspectral testing data of the seeds to be tested;

the preprocessing module 22: the hyperspectral data processing system is used for preprocessing the hyperspectral test data to obtain hyperspectral characteristic data;

spectral information processing and fusion module 23: the spectrum first-order differential model is used for extracting the characteristics of the hyperspectral characteristic data to determine key spectrum characteristics;

the state determination module 24: and the method is used for calculating according to the key wave band characteristics and a regression analysis algorithm to determine the vitality state of the seeds.

In the embodiment of the invention, relevance is established by establishing hyperspectral data and the root length after germination, key spectral information of hyperspectrum is obtained by adopting first-order difference, and then the root length prediction of seedlings is determined by adopting regression analysis; the method can realize rapid, nondestructive and high-precision seed viability detection.

EXAMPLE III

Referring to fig. 8, fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the disclosure. The electronic device may be a computer, a server, or the like, and may also be an intelligent device such as a mobile phone, a tablet computer, a monitoring terminal, or the like, and an image acquisition device having a processing function. As shown in fig. 8, the electronic device may include:

a memory 510 storing executable program code;

a processor 520 coupled to the memory 510;

wherein, the processor 520 calls the executable program code stored in the memory 510 to execute part or all of the steps of the method for predicting germination of sweet corn seeds in the first embodiment.

The embodiment of the invention discloses a computer-readable storage medium which stores a computer program, wherein the computer program enables a computer to execute part or all of the steps in the sweet corn seed germination prediction method in the first embodiment.

The embodiment of the invention also discloses a computer program product, wherein when the computer program product runs on a computer, the computer is caused to execute part or all of the steps in the sweet corn seed germination prediction method in the first embodiment.

The embodiment of the invention also discloses an application publishing platform, wherein the application publishing platform is used for publishing the computer program product, and when the computer program product runs on a computer, the computer is enabled to execute part or all of the steps in the sweet corn seed germination prediction method in the first embodiment.

In various embodiments of the present invention, it should be understood that the sequence numbers of the processes do not mean the execution sequence necessarily in order, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated units, if implemented as software functional units and sold or used as a stand-alone product, may be stored in a computer accessible memory. Based on such understanding, the technical solution of the present invention, which is a part of or contributes to the prior art in essence, or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several requests for causing a computer device (which may be a personal computer, a server, a network device, or the like, and may specifically be a processor in the computer device) to execute part or all of the steps of the method according to the embodiments of the present invention.

In the embodiments provided herein, it should be understood that "B corresponding to a" means that B is associated with a from which B can be determined. It should also be understood, however, that determining B from a does not mean determining B from a alone, but may also be determined from a and/or other information.

Those of ordinary skill in the art will appreciate that some or all of the steps of the methods of the embodiments may be implemented by instructions associated with hardware via a program, and the program may be stored in a computer-readable storage medium, where the storage medium includes Read-Only Memory (ROM), Random Access Memory (RAM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), One-time Programmable Read-Only Memory (OTPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM), or other Memory, such as a magnetic disk, a, A tape memory, or any other medium readable by a computer that can be used to carry or store data.

The sweet corn seed germination prediction method, the sweet corn seed germination prediction device, the electronic equipment and the storage medium disclosed by the embodiment of the invention are described in detail, a specific example is applied in the method to explain the principle and the implementation mode of the invention, and the description of the embodiment is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. A sweet corn seed germination prediction method is characterized by comprising the following steps:

acquiring hyperspectral test data of seeds to be tested;

preprocessing the hyperspectral test data to obtain hyperspectral characteristic data;

performing characteristic acquisition on the hyperspectral characteristic data by adopting a spectrum first-order differential model to determine corresponding key waveband characteristics;

2. The sweet corn seed germination prediction method of claim 1, wherein preprocessing the hyperspectral test data to obtain hyperspectral characteristic data comprises:

3. The sweet corn seed germination prediction method according to claim 1, wherein the hyperspectral test data is hyperspectral data in a spectral range of 400nm to 1000 nm; the hyperspectral test data is one of hyperspectral data of the embryo surface of the seed or hyperspectral data of the embryo breast surface of the seed.

4. The sweet corn seed germination prediction method of claim 3, wherein the performing feature extraction on the hyperspectral feature data by using the spectral first order difference information acquisition model to determine corresponding key spectral features comprises:

carrying out characteristic wave band division on the hyperspectral characteristic data according to a preset number of spectral bands;

assume that the original data sequence is y_i(1 ≦ i ≦ n), and the variable y is defined on a non-negative integer value; when the variable i sequentially takes non-negative integers, and when i is changed from k to k +1, the change amount of the dependent variable is as follows:

△y_k＝y_k+1-y_k

△y_kreferred to as the first difference of the function at point k;

spectral key feature information is determined based on the spectral data and its first order difference information.

5. The method for predicting germination of sweet corn seeds as claimed in claim 4, wherein the number of the variables of the original spectrum band data is preset to be 220.

6. The sweet corn seed germination prediction method of claim 1, wherein the calculating based on the key band features and a regression analysis algorithm to determine seed vigor status comprises:

and (3) carrying out regression analysis on the key characteristic wave band by adopting a principal component regression algorithm, a partial least square regression algorithm and a support vector regression algorithm to determine the vitality state of the seeds, wherein the state characteristic of reflecting the vitality of the seeds is the seedling root length.

7. The method of predicting germination of a sweet corn seed of claim 6, wherein the kernel function selected by the support vector regression algorithm is a Gaussian radial basis function; the gaussian radial basis function is:

k(x,y)＝exp(-||x-y||²k (x, y), where σ is selected to be 50.

8. A sweet corn seed germination prediction device, comprising:

a preprocessing module: the hyperspectral test data acquisition unit is used for preprocessing the hyperspectral test data to obtain hyperspectral characteristic data;

a band selection module: the hyperspectral characteristic data are selected and analyzed by adopting a waveband selection model to determine corresponding key waveband characteristics;

9. An electronic device, comprising: a memory storing executable program code; a processor coupled with the memory; the processor calls the executable program code stored in the memory for performing the sweet corn seed germination prediction method of any one of claims 1 to 7.

10. A computer readable storage medium storing a computer program, wherein the computer program causes a computer to perform the sweet corn seed germination prediction method of any one of claims 1 to 7.