CN110082298B - Hyperspectral image-based wheat variety gibberellic disease comprehensive resistance identification method - Google Patents
Hyperspectral image-based wheat variety gibberellic disease comprehensive resistance identification method Download PDFInfo
- Publication number
- CN110082298B CN110082298B CN201910401015.1A CN201910401015A CN110082298B CN 110082298 B CN110082298 B CN 110082298B CN 201910401015 A CN201910401015 A CN 201910401015A CN 110082298 B CN110082298 B CN 110082298B
- Authority
- CN
- China
- Prior art keywords
- wheat
- disease
- resistance
- scab
- variety
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/27—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands using photo-electric detection ; circuits for computing concentration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2201/00—Features of devices classified in G01N21/00
- G01N2201/12—Circuits of general importance; Signal processing
- G01N2201/129—Using chemometrical methods
- G01N2201/1296—Using chemometrical methods using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Immunology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Analytical Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Pathology (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biochemistry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
- Breeding Of Plants And Reproduction By Means Of Culturing (AREA)
Abstract
The invention discloses a method for identifying the comprehensive resistance of wheat variety to gibberellic disease based on hyperspectral images, which takes wheat ears infected with the scab germs of different varieties as research objects, comprehensively applies knowledge in various fields such as hyperspectral imaging technology, spectroscopy, time series analysis, deep learning and the like, and provides the following steps: a resistance identification method aiming at a wheat variety in a non-morbidity infection period, a resistance identification method aiming at a wheat variety after morbidity and a comprehensive resistance identification method aiming at wheat variety scab. The invention overcomes the defects of difficult resistance detection in the infection period, low speed and low precision in manual identification operation, complex program and high cost in a chemical method, provides a faster and more accurate method for identifying the resistance of the wheat variety to the gibberellic disease, and can identify the resistance of the wheat variety in the period when the wheat is infected with the gibberellic disease but does not develop the disease.
Description
Technical Field
The invention relates to the field of wheat scab detection and resistance identification, hyperspectral imaging technology, time sequence analysis and deep learning algorithm, in particular to a method for comprehensively identifying the scab resistance of a wheat variety based on a hyperspectral image.
Background
Wheat is one of three grains in the world, and the problem of wheat diseases is a key point of attention of people. The wheat scab is a high-incidence mycosis which has the greatest threat to wheat production and is also called red wheat head, red head miasma, rotten wheat head and wheat head blight, wheat grains infected with the wheat scab can generate mycotoxin (slow flying; the influence of different infection periods on wheat scab generation and DON accumulation in the grains) mainly comprising deoxynivalenol in the infection process, the serious loss of yield can be caused, and the mycotoxin secreted by the wheat scab can seriously harm human health, so that the food safety problem is caused. The wheat yield loss caused by the gibberellic disease is heavy, and the wheat yield is reduced by 50% or even is out of production in the epidemic years of the gibberellic disease. Moreover, wheat scab is currently in an incurable stage and once infected with scab, incurable severe losses are certainly incurred. At present, even if the pesticide is used, the treatment effect cannot be guaranteed, the pesticide is not environment-friendly, and the environment is greatly polluted. Therefore, it is a serious problem to select a wheat variety resistant to scab for breeding. The identification of the resistance to gibberellic disease of wheat variety is an important link in wheat breeding.
At the present stage, the identification of the wheat scab resistance is mainly reflected by the head disease rate and the small head disease rate of wheat infected with scab, and the detection of the head disease rate and the small head disease rate is mainly observed and counted by professional plant protection personnel through manual naked eyes. In addition to manual detection, various chemical and biological monitoring methods have been developed at home and abroad for detecting DON toxins produced in wheat grains after diseases, and in the research on wheat flour, wheat grains need to be ground into powder by the chemical methods, such as Thin Layer Chromatography (TLC), High Performance Liquid Chromatography (HPLC), and the like. In addition to a complex chemical method, researchers at the present stage use a hyperspectral technology to carry out grading diagnosis on wheat scab, and finally find that a local threshold segmentation method has the best grading effect and the grading accuracy rate reaches 90.78% by extracting characteristic wave bands of two times of principal component analysis and adopting different threshold segmentation methods. However, the local threshold segmentation method has disadvantages, the method trains the model for detecting the disease level of wheat scab in a machine learning manner, the accuracy and the simplicity are not as good as deep learning, the local threshold segmentation method may process noise problems such as wheat awn and the like into different regions caused by lesion positions, and data have certain errors, so that the method cannot be completely applied to a data set with higher complexity only by using a typical classification algorithm. The deep network in the deep neural network can automatically extract effective spectral features, so that end-to-end model construction is realized. Most importantly, studies on wheat scab resistance are compared by the ear rate or grading condition of wheat after disease occurrence, and no method is available for detecting the condition of infection of germs but not the disease occurrence.
Based on the problems, the invention provides a method for identifying the comprehensive resistance of the gibberellic disease of the wheat variety based on a hyperspectral image. Overcomes the defects of difficult resistance detection in the infection period, low speed and precision in manual identification operation, complex program and high cost in a chemical method, and the like, not only provides a faster and more accurate method for the resistance identification of the wheat scab of the variety, but also can carry out the resistance identification on the wheat variety in the period when the wheat is infected with the fusarium graminearum but is not sick.
Disclosure of Invention
The invention aims to provide a method for identifying the comprehensive resistance of the gibberellic disease of a wheat variety based on a hyperspectral image aiming at the problems in the prior art; the method fills the blank of resistance detection of wheat scab infection in the non-morbidity period, and the defects of low speed, low precision, complex program, high cost and the like in a chemical method in the artificial resistance identification operation after morbidity, covers the initial morbidity time detection, the spectrum and image characteristic difference analysis and the automatic grading detection of the grade after morbidity of the wheat scab infection in the non-morbidity period, comprehensively considers the initial morbidity time of the infection in the non-morbidity period and the average severity of the morbidity, and can comprehensively, quickly and accurately realize the comprehensive resistance identification of the wheat variety scab.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a method for identifying the comprehensive resistance of wheat variety to scab based on hyperspectral images is characterized by comprising the steps of identifying the resistance of wheat varieties in the infection non-morbidity period, identifying the resistance of wheat varieties after morbidity and identifying the comprehensive resistance of wheat varieties to scab based on the two results; the specific identification method comprises the following steps:
a. collecting wheat samples: selecting samples from wheat to be detected, and marking corresponding marks according to varieties;
b. collecting hyperspectral images of wheat ears: imaging the wheat ear to be detected by utilizing a hyperspectral imaging system, and carrying out resistance comparison in the non-morbidity period of wheat scab infection and establishment of a wheat scab grade detection model after morbidity through the obtained image;
c. comparison of resistance in the non-diseased stage of wheat scab infection: the method comprises the steps of detecting the initial attack time of wheat scab of different varieties and researching the resistance of wheat scab of different varieties in the same infection period; the method comprises the following steps of detecting the initial attack time of the gibberellic disease of different varieties of wheat, wherein the detection is carried out by exploring the rule and the characteristic that the spectrum changes along with the infection time after the wheat ear is infected with germs, analyzing a time sequence key point by adopting a dynamic time warping clustering algorithm based on the time sequence characteristic, and detecting the initial attack time of the gibberellic disease of different varieties of wheat based on the time sequence hyperspectral so as to compare the gibberellic disease resistance of different varieties of wheat; the study on the resistance to the wheat scab of different varieties of wheat in the same infection period is to find out the spectrum characteristics and the image characteristic variation differences of wheat ears in the same infection period of different varieties and analyze and compare the resistance to the wheat scab of different varieties based on the spectrum characteristic differences and the image characteristic differences;
d. establishing a wheat scab grade detection model after disease incidence: performing ROI extraction, pretreatment and characteristic waveband image selection on a wheat ear sample subjected to complete morbidity, establishing a wheat scab grade detection model based on a hyperspectral image by combining an RCNN-VGG16 algorithm, realizing target detection and labeling of ill spikelets and normal spikes, and comparing the scab resistance of wheat varieties according to the ill average severity of the wheat scabs of different varieties;
e. and (3) identifying the comprehensive resistance of wheat variety scab: and (d) identifying the wheat resistance standard according to the average severity of the scab of the wheat in national agricultural industry standard NY/T1443.4-2007 by comparing the initial disease time (the result obtained in the step c) and the average severity of the disease after the disease of the wheat of different varieties (the result obtained in the step d), and specifically classifying the wheat varieties into comprehensive resistance disease-resistant types, comprehensive resistance susceptible types and comprehensive resistance susceptible types by combining the initial disease time.
Further, in the step b, the hyperspectral imaging system comprises a CCD camera, an Imspector spectrometer, a lens, a 21V/150W linear halogen lamp light source, a dark box and a PC; the hyperspectral imaging wave band range is 358-1021 nm.
Further, in the steps c and d, performing black and white correction on the system, collecting hyperspectral images of the wheat ears after inoculation every 24 hours until the wheat ears are completely attacked in 21 days according to detection research of the initial attack time of gibberellic disease of different varieties of wheat, and marking according to varieties; the method comprises the steps of collecting hyperspectral images of different varieties 3 days, 7 days, 14 days and 21 days after inoculation of germs aiming at analysis and research of variation differences of spectral features and image features of wheat ears of different varieties in the same infection period, and marking infection time and varieties; the establishment of a wheat scab grade detection model after disease incidence is to collect hyperspectral images of wheat ears which are completely attacked 21 days after all samples of all varieties are collected.
Further, the hyperspectral images of wheat ears of different varieties collected every 24 hours are used for processing research, the hyperspectral images from the same sample are used for constructing a time sequence hyperspectral image, corresponding to a scab area when the disease is completely developed at last, the region of interest is extracted by envi software to obtain a time sequence spectrum, the spectrum is preprocessed by a convolution smoothing algorithm, the feature extraction is carried out on the time sequence spectrum by a continuous projection algorithm, the time sequence spectrum data is subjected to cluster analysis by a dynamic time warping clustering algorithm, time sequence key point extraction is carried out, and the initial point of the disease development stage is diagnosed; based on the obtained detection results of the scab initial attack time of different varieties of wheat with time sequence hyperspectrum, the average initial attack time of the varieties is calculated, so as to compare the scab resistance of different varieties of wheat infected with the non-attack period.
Further, hyperspectral images of different collected varieties in 3 days, 7 days, 14 days and 21 days after inoculation of germs are used for processing and analyzing, hyperspectral images from the same sample are extracted from the interested region of the whole wheat ear by envi software, average calculation is carried out on the interested region pixel points to obtain average reflectivity spectrums of different varieties of wheat in the same infection period after inoculation, spectrograms are observed, and spectrum differences are compared on the spectrograms.
Furthermore, the differences of the area, the length and the gray value of an infected area on a hyperspectral image under the characteristic wave band are observed, and the resistance of wheat scab of different varieties in the period without infection is compared through the obtained spectral difference and the image difference.
Further, in the step d, performing whole wheat ear ROI extraction on the hyperspectral image after complete morbidity (after 21 days of inoculation) by utilizing envi software, preprocessing the spectrum by utilizing an SNV algorithm, stably correlating chlorophyll fluorescence parameters of the wheat in a visible light range of 520-680nm according to a wheat chlorophyll reflection spectrum, observing the hyperspectral image at about 650nm by naked eyes in the range, selecting a wheat image of the wave band, marking a disease-free ear as normal by using a marking software labelImg, training by using a fast RCNN algorithm based on five convolutional neural network models after a disease-present ear is found as sick, and establishing a wheat ear gibberellic disease infection grade identification model by utilizing an RCNN-VGG16 algorithm based on the wave band.
Furthermore, the selected wheat picture is input into a convolutional neural network to obtain feature mapping, the feature mapping is input into a region suggestion network to obtain feature information of a candidate frame, a classifier is used for judging whether the candidate frame belongs to a diseased spikelet or a non-diseased spikelet, and then comparison with a result of manual measurement is carried out to optimize an algorithm; and calculating the average severity of diseases of different varieties according to the grade of the head scab to determine the scab resistance of different varieties after the diseases occur.
Further, in the step e, a method for identifying the comprehensive resistance of the gibberellic disease of the wheat variety is provided by combining the initial morbidity time obtained in the step c and the average severity of diseases after the wheat variety is ill obtained in the step d: namely when the initial onset time of the wheat variety is 15-21 days and the average severity after onset is more than 0 and less than 2.0, the wheat variety is a comprehensive resistance disease-resistant variety; when the initial disease onset time of the wheat variety is 10-15 days and the sum of the average severity after the disease onset is more than 2.0 and less than 3.0, the wheat variety is the comprehensive resistance resistant variety; when the initial onset time of the wheat variety is 5-10 days and the average severity after onset is more than 3.0 and less than 3.5, the wheat variety is the comprehensive resistant susceptible variety, and when the initial onset time of the wheat variety is 0-5 days and the average severity after onset is more than 3.5, the wheat variety is the comprehensive resistant susceptible variety.
Further, the average severity after the onset of the wheat variety is calculated by the following formula:
wherein, a: the number of first-grade samples of the variety; b: the number of second-level samples of the variety; c: the number of the three-level samples of the variety; d: the number of the four-level samples of the variety; m: the total number of samples of the variety.
Compared with the prior art, the invention has the following advantages:
(1) the research on the gibberellic disease resistance of the wheat variety can effectively solve the breeding problem of wheat, and select a variety with higher resistance for breeding, thereby reducing the hidden danger of the wheat food safety.
(2) Compared with the existing manual and chemical detection methods, the method for detecting the resistance of the wheat variety to the gibberellic disease can be nondestructively, rapidly and simply carried out by utilizing the research on the hyperspectral image.
(3) The method for researching the wheat scab infection non-morbidity stage by utilizing the hyperspectral image provides an effective means for researching the wheat variety scab resistance, and makes up the blank that the variety resistance can not be detected in the wheat scab infection non-morbidity stage.
(4) The method has the advantages that the grade of wheat scab after disease attack is detected, a large number of data sets can be processed by combining with the hyperspectral characteristic waveband image and utilizing a deep learning algorithm, end-to-end model construction is achieved, and the detection accuracy can be improved.
(5) A more comprehensive identification method for the gibberellic disease comprehensive resistance of the wheat variety is firstly provided by combining the onset time of the wheat scab infection in the non-morbidity stage and the average severity of diseases after the disease onset, which are obtained by utilizing the hyperspectral image.
In summary, the invention relates to a hyperspectral imaging technology, spectroscopy, time series analysis and deep learning algorithm and related technologies in the field thereof, and mainly aims to solve the problems that the detection of the wheat ear disease rate at the present stage is high in detection cost, time-consuming and labor-consuming, complex in procedure, and only capable of detecting a completely diseased wheat ear sample. The invention provides a hyperspectral image-based method for identifying the comprehensive resistance of wheat varieties, which is characterized in that the detection of the initial attack time of hyperspectral images of different varieties of wheat in the non-attack period, the analysis of the spectral characteristics and the image characteristics of the wheat in the same infection period of different varieties and the detection model of the wheat scab grade after attack based on a hyperspectral image by utilizing a deep learning algorithm are carried out, the detection of the attack initial time of the wheat scab of different varieties, the comparison of the spectral characteristics and the image characteristics of the wheat in the same infection period of different varieties and the automatic grading of the grade after attack are realized, and the identification of the comprehensive resistance of the wheat varieties to the scab is realized by combining the average severity of the attack calculated by the attack initial time of the wheat scab and the grading result of.
Drawings
FIG. 1 is a general technical roadmap for the present invention.
FIG. 2 is a technical route chart for detecting the onset time of wheat.
FIG. 3 is a roadmap for wheat scab resistance studies of different varieties at the same infection stage.
FIG. 4 is a technical scheme of an automatic wheat scab grading test model.
FIG. 5 is a graph showing the results of wheat ears of various grades of head blight; in fig. 5(a), 5(b), 5(c), 5(d), 5(e), and 5(e) are respectively a grade 0 ear result chart, a grade 1 ear result chart, a grade 2 ear result chart, a grade 3 ear result chart, and a grade 4 ear result chart, respectively.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments.
(1) Wheat sample collection and processing
Sample selection: samples in the experiment are all from experimental farms in Liuhe district of Nanjing City, four kinds of healthy wheat in flowering stages of Shuibai, Zhenmai 168, Huaimai and Luomai 22 are expected in the farms, the four kinds of wheat ears are respectively cultivated, and corresponding marks are made according to varieties.
culturing the gibberellic disease by using a potato glucose agar (PDA) culture medium, namely cleaning potatoes, peeling, cutting into 400g blocks, heating in boiling water for 5min until the potatoes are soft, filtering, pouring filtrate into a pot for heating, adding agar, stirring, adding glucose, continuously heating for 5min, sealing and storing, separating and purifying the bacteria, namely obtaining diseased wheat grains, treating the wheat grains by mercuric chloride-70% alcohol-sterilized water, adding 0.3% gentamycin into the PDA culture medium, inoculating the grains into the culture medium, manually culturing at constant temperature, picking up and purifying the mycelia from the edges of the colonies after 48h, transferring the mycelia to a slant of a PDA test tube, growing in a counter-environment, and growing the mycelia for several daysFull PDA, in preparation of 1X 105Single/ml conidia suspension.
Inoculation experiment: and (3) completely inoculating all the wheat ear samples collected in the first step, injecting the spore suspension into the single flowers at the same part of the wheat ear by using an injector, and covering a transparent plastic bag for moisturizing.
(2) Hyperspectral image acquisition of wheat ears
The hyperspectral imaging system mainly comprises main components such as a CCD camera (Imprix, America), an imager spectrometer (Specim, Finland), a lens (Schneider, Germany), a 21V/150W linear halogen lamp light source (Ill. ), a dark box, a PC and the like. The hyperspectral imaging wave band range is 358-1021 nm. In order to obtain clear image information, parameters of a test system are adjusted for multiple times, the relation between factors such as a measured spectrum region and a collection environment (light source, time, temperature and the like) and single ear information is researched, and the spectrum region capable of effectively bearing characteristic information of a sample, and a proper spectrum measurement mode and measurement parameters are established.
And performing black-and-white correction on the system, and collecting hyperspectral images of the wheat ears after inoculation every 24 hours until the wheat ears completely attack after 21 days according to detection research of the initial attack time of the scab of the wheat of different varieties, and marking according to the varieties. The method aims at the analysis and research of the variation difference of the spectral characteristics and the image characteristics of the wheat ears in the same infection period of different varieties, collects the hyperspectral images of the different varieties 3 days, 7 days, 14 days and 21 days after the inoculation of germs, and marks the infection time and the varieties. Aiming at the research establishment of a wheat scab grade detection model after disease occurrence, hyperspectral images of wheat ears which are completely diseased after all samples 21 of all varieties are collected.
(3) Artificial measurement of scab infection wheat ear rate
According to the standard GB/T15796-2011, the collected wheat ear samples are counted by combining the manual identification rule with the agricultural production experience, and the samples are classified according to the infection first-class (proportion < 25%), the second-class (proportion < 25% < 50%), the third-class (proportion < 75%) and the fourth-class (proportion > 75%) according to the proportion of the infected ear to the total ear of the wheat ear.
(4) Detection model for infecting scab resistance in different periods based on hyperspectral images
the method comprises the steps of establishing initial attack time models of wheat scabs of different varieties, utilizing hyperspectral images of wheat ears of different varieties collected every 24 hours to process and study, constructing time sequence hyperspectral images from the same sample, utilizing envi software to extract regions of interest (ROI) corresponding to lesion areas when complete attack occurs finally to obtain time sequence spectrums, preprocessing the spectrums by using a convolution smoothing algorithm (SG), extracting features of the time sequence spectrums by using a continuous projection algorithm (SPA), performing cluster analysis by using Dynamic Time Warping (DTW) clustering algorithm time sequence spectrum data, extracting time sequence key points, diagnosing initial attack points of the attack period, calculating average initial attack time of the varieties based on detection results of the initial attack time of the wheat scabs of the different varieties of the time sequence hyperspectrum, and comparing the scabs resistance of the wheat of the different varieties, wherein the specific technical route diagram is shown in figure 2.
the hyperspectral images of the wheat ears in the same infection period of different varieties are used for processing and researching the spectral characteristics and the image characteristic variation difference of the wheat ears in the same infection period of different varieties, the collected hyperspectral images of the wheat ears in 3 days, 7 days, 14 days and 21 days after the inoculation of germs are used for extracting a region of interest (ROI) of the whole wheat ear by using envi software, and the average reflectance spectrum of the wheat ears in the same infection period of different varieties after the inoculation is obtained by average calculation of pixel points of the region of interest (ROI). on an observation spectrogram, the spectral reflectance difference of the wheat ears in the same infection period within the wavelength range of different varieties is obtained, because of the infection of gibberellic disease, the spectral information is related to the contraction vibration and frequency doubling of the C-H, O-H, N-H key which is the main structural component of organic molecules such as protein, water, starch and the like, therefore, the spectral difference can be compared on the spectrogram, the 3-level frequency doubling of the C-H, the 2-level frequency of the C-H, the area of the approximate 600nm and 680nm, the area of other pigments and the spectral difference of the wavelength of the anthocyanin, the observation area of the wheat ears, the peak value of the wheat ears, and the wavelength difference of the wheat are obtained by comparison, the spectral reflectance difference of the experimental analysis of the peak value of the experimental spectrum, the experimental result of the experimental analysis of the.
(5) Wheat scab grade detection model based on hyperspectral image
Extracting the ROI of the whole wheat ear from the hyperspectral image of the wheat ear sample after complete morbidity by using envi software, preprocessing the spectrum by using an SNV algorithm, stably correlating the chlorophyll reflectance spectrum of the wheat with chlorophyll fluorescence parameters in the range of 520-680nm of visible light, observing the hyperspectral image of about 650nm by naked eyes in the range, selecting a wheat picture of the waveband, labeling a disease-free ear as normal by using label software labelImg, training by using a FasterRCNN algorithm based on five convolutional neural network models after the disease-free ear is sick, and establishing a wheat ear gibberellic disease infection grade identification model by using an RCNN-VGG16 algorithm based on the waveband. Inputting the picture into a convolutional neural network to obtain feature mapping, inputting a feature mapping input region suggestion network (RPN) to obtain feature information of a candidate frame, judging whether the candidate frame belongs to a diseased spikelet or a non-diseased spikelet through a classifier, comparing the result with a result of manual measurement, and optimizing an algorithm. The trained model is tested by using a test set consisting of 50 infected wheat pictures containing 5 grades, the mAP value of the finally obtained model is 0.857, the FPS (frame number of processed images per second) is 18, and the grading total accuracy rate reaches 92%. And calculating the average severity of diseases of different varieties according to the grade of the head scab to determine the scab resistance of different varieties. The specific technical scheme is shown in figure 4. The head results of the grades of head of gibberellic disease are shown in FIG. 5.
(6) Method for identifying comprehensive resistance of wheat variety to scab
The detection of the onset time of the wheat scab infection non-onset period is combined, and the average severity of the disease after the disease onset is provided with a comprehensive resistance identification method for the scab of the wheat variety, wherein the method specifically requires that the wheat variety with the onset initial time of 15-21 days and the average severity of more than 0 and less than 2.0 after the disease onset is a comprehensive resistance disease-resistant variety; when the initial disease onset time of the wheat variety is 10-15 days and the sum of the average severity after the disease onset is more than 2.0 and less than 3.0, the wheat variety is the comprehensive resistance resistant variety; when the initial onset time of the wheat variety is 5-10 days and the average severity after onset is more than 3.0 and less than 3.5, the wheat variety is the comprehensive resistant susceptible variety, and when the initial onset time of the wheat variety is 0-5 days and the average severity after onset is more than 3.5, the wheat variety is the comprehensive resistant susceptible variety. As shown in Table 1 below
| 15<Time<21&0<Mean severity<2.0 | Comprehensive resistance disease-resistant variety |
| 10<Time<15&2.0<Mean severity<3.0 | Resistant variety in comprehensive resistance |
| 5<Time<10&3.0<Mean severity<3.5 | Synthetic resistant and |
| 0<Time<5&Mean severity>＝3.5 | Comprehensive resistant susceptible variety |
TABLE 1 evaluation criteria for comprehensive resistance identification of wheat scab
(a: the number of the first-class samples of the variety, b: the number of the second-class samples of the variety;
c: the number of the three-level samples of the variety; d: the number of the four-level samples of the variety; m: the total sample number of the variety; )
The foregoing illustrates and describes the principles, general features, and advantages of the present invention. It should be understood by those skilled in the art that the above embodiments do not limit the scope of the present invention in any way, and all technical solutions obtained by using equivalent substitution methods fall within the scope of the present invention. The parts not involved in the present invention are the same as or can be implemented using the prior art.
Claims (10)
1. A method for identifying the comprehensive resistance of wheat variety scab based on hyperspectral images is characterized by comprising the resistance identification of wheat varieties in the infection non-morbidity period, the resistance identification of wheat varieties after morbidity and the comprehensive resistance identification of wheat variety scab based on the two research results; the specific identification method comprises the following steps:
a. collecting wheat samples: selecting samples from wheat to be detected, and marking corresponding marks according to varieties;
b. collecting hyperspectral images of wheat ears: imaging the wheat ear to be detected by utilizing a hyperspectral imaging system, and carrying out resistance comparison in the non-morbidity period of wheat scab infection and establishment of a wheat scab grade detection model after morbidity through the obtained image;
the resistance comparison of the wheat scab infection at the non-morbidity stage is carried out: the method comprises the steps of detecting the initial attack time of wheat scab of different varieties and researching the resistance of wheat scab of different varieties in the same infection period; the method comprises the following steps of detecting the initial attack time of the gibberellic disease of different varieties of wheat, wherein the detection is carried out by exploring the rule and the characteristic that the spectrum changes along with the infection time after the wheat ear is infected with germs, analyzing a time sequence key point by adopting a dynamic time warping clustering algorithm based on the time sequence characteristic, and detecting the initial attack time of the gibberellic disease of different varieties of wheat based on the time sequence hyperspectral so as to compare the gibberellic disease resistance of different varieties of wheat; the study on the resistance to the wheat scab of different varieties of wheat in the same infection period is to find out the spectrum characteristics and the image characteristic variation differences of wheat ears in the same infection period of different varieties and analyze and compare the resistance to the wheat scab of different varieties based on the spectrum characteristic differences and the image characteristic differences;
establishing a wheat scab grade detection model after disease incidence: performing ROI extraction, pretreatment and characteristic waveband image selection on a wheat ear sample subjected to complete morbidity, establishing a wheat scab grade detection model based on a hyperspectral image by combining an RCNN-VGG16 algorithm, realizing target detection and labeling of ill spikelets and normal spikes, and comparing the scab resistance of wheat varieties according to the ill average severity of the wheat scabs of different varieties;
c. and (3) identifying the comprehensive resistance of wheat variety scab: through comparison of initial disease attack time and average severity of diseases after the disease attack of different varieties of wheat, the resistance standard of the wheat is identified according to the average severity of the disease attack of wheat scab of national agricultural industry standard NY/T1443.4-2007, and different varieties of wheat are specifically divided into comprehensive resistance disease-resistant types, comprehensive resistance susceptible types and comprehensive resistance susceptible types by combining the initial disease attack time.
2. The method for identifying the comprehensive resistance to gibberellic disease of wheat variety based on hyperspectral image as claimed in claim 1, wherein in step b, the hyperspectral imaging system comprises CCD camera, imager spectrometer, lens, 21V/150W linear halogen lamp light source, dark box and PC; the hyperspectral imaging wave band range is 358-1021 nm.
3. The method for identifying the comprehensive resistance of the gibberellic disease of the wheat variety based on the hyperspectral image as claimed in claim 1, wherein in the step of performing resistance comparison in the non-morbidity period of wheat scab infection and the step of establishing the wheat scab grade detection model after morbidity, black and white correction is performed on the system, detection research aiming at the initial morbidity time of the gibberellic disease of different varieties of wheat is performed, a hyperspectral image of an inoculated wheat ear is collected every 24 hours, the complete morbidity is stopped after 21 days, and the hyperspectral image is marked according to varieties; the method comprises the steps of collecting hyperspectral images of different varieties 3 days, 7 days, 14 days and 21 days after inoculation of germs aiming at analysis and research of variation differences of spectral features and image features of wheat ears of different varieties in the same infection period, and marking infection time and varieties; the establishment of a wheat scab grade detection model after disease incidence is to collect hyperspectral images of wheat ears which are completely attacked 21 days after all samples of all varieties are collected.
4. The method for identifying the comprehensive resistance of the gibberellic disease of the wheat variety based on the hyperspectral image is characterized in that aiming at the detection research of the initial attack time of the gibberellic disease of different varieties of wheat, the hyperspectral images of different varieties of wheat ears collected every 24 hours are utilized for processing research, the hyperspectral images from the same sample are used for constructing a time sequence hyperspectral image, corresponding to a scab area when the disease is completely finally attacked, the region of interest is extracted by utilizing envi software to obtain a time sequence spectrum, the spectrum is preprocessed by utilizing a convolution smoothing algorithm, the feature extraction is carried out on the time sequence spectrum by utilizing a continuous projection algorithm, the time sequence spectrum data is subjected to cluster analysis by utilizing a dynamic time warping clustering algorithm, the time sequence key point extraction is carried out, and the initial point of the disease attack stage is diagnosed; based on the obtained detection results of the scab initial attack time of different varieties of wheat with time sequence hyperspectrum, the average initial attack time of the varieties is calculated, so as to compare the scab resistance of different varieties of wheat infected with the non-attack period.
5. The method for identifying the gibberellic disease comprehensive resistance of the wheat variety based on the hyperspectral image as claimed in claim 1 or 3 is characterized in that hyperspectral image analysis and image feature difference analysis research of different varieties in the same infection period is carried out, the collected hyperspectral images of different varieties in 3 days, 7 days, 14 days and 21 days after germ inoculation are used for processing and analyzing, the hyperspectral image from the same sample is extracted from the interested area of the whole wheat ear by envi software, the average reflectance spectrum of the inoculated wheat variety in the same infection period is obtained by average calculation of the interested area of pixel points, the spectrogram is observed, and the spectrum difference is compared on the spectrogram.
6. The method for identifying the wheat variety scab comprehensive resistance based on the hyperspectral images as claimed in claim 5, wherein the differences of the area, length and gray value of the infected area on the hyperspectral image under the characteristic wave band are observed, and the obtained spectral difference and image difference are used for comparing the scab resistance of different varieties of wheat in the same infection period.
7. The method for identifying the comprehensive resistance to the gibberellic disease of the wheat variety based on the hyperspectral image as claimed in claim 1, wherein in the step of establishing the post-morbidity wheat scab grade detection model, the specific method for establishing the post-morbidity wheat scab grade detection model comprises the following steps: extracting the ROI (region of interest) of the whole wheat ear from the hyperspectral image of the wheat ear sample after complete outbreak of disease by using envi software, preprocessing the spectrum by using an SNV (noise-plus-volt) algorithm, observing the hyperspectral image about 650nm by naked eyes, selecting a wheat image of the wave band, marking a disease-free ear as normal by using label software labelImg, training by using a fast RCNN (Radar Cross-coupled neural network) algorithm based on five convolutional neural network models after the disease-free ear is sick, and establishing a wheat ear gibberellic disease infection grade identification model by using an RCNN-VGG16 algorithm based on the wave band.
8. The method for identifying the gibberellic disease comprehensive resistance of the wheat variety based on the hyperspectral image as claimed in claim 7, wherein in the training of the fast RCNN algorithm, the selected wheat picture is input into a convolutional neural network to obtain feature mapping, the feature mapping is input into a regional suggestion network to obtain feature information of a candidate frame, and the candidate frame is judged to belong to a diseased spikelet or a non-diseased spikelet through a classifier, and then compared with the result of manual measurement to optimize the algorithm; and calculating the average severity of diseases of different varieties according to the grade of the head scab to determine the scab resistance of different varieties after the diseases occur.
9. The method for identifying the comprehensive resistance to gibberellic disease of a wheat variety based on a hyperspectral image as claimed in claim 1, wherein in the step c, the comprehensive resistance to gibberellic disease of the wheat variety is identified by integrating the detection result of the initial disease attack time of the step for comparing the resistance of the wheat scab infection in the non-disease attack period with the detection result of the severity of the disease after the disease attack of the step established by the detection model of the grade of the wheat scab after the disease, namely, the wheat variety with the initial disease attack time of 15-21 days and the average severity of the disease after the disease of more than 0 and less than 2.0 is an integrated resistance and disease-resistant variety; when the initial disease onset time of the wheat variety is 10-15 days and the sum of the average severity after the disease onset is more than 2.0 and less than 3.0, the wheat variety is the comprehensive resistance resistant variety; when the initial onset time of the wheat variety is 5-10 days and the average severity after onset is more than 3.0 and less than 3.5, the wheat variety is the comprehensive resistant susceptible variety, and when the initial onset time of the wheat variety is 0-5 days and the average severity after onset is more than 3.5, the wheat variety is the comprehensive resistant susceptible variety.
10. The method for identifying the comprehensive resistance to gibberellic disease of the wheat variety based on the hyperspectral image as claimed in claim 1 or 9, wherein the average severity of the wheat variety after the onset of disease is calculated by the following formula:
wherein, a: the number of first-grade samples of the variety; b: the number of second-level samples of the variety; c: the number of the three-level samples of the variety; d: the number of the four-level samples of the variety; m: the total number of samples of the variety.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910401015.1A CN110082298B (en) | 2019-05-15 | 2019-05-15 | Hyperspectral image-based wheat variety gibberellic disease comprehensive resistance identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910401015.1A CN110082298B (en) | 2019-05-15 | 2019-05-15 | Hyperspectral image-based wheat variety gibberellic disease comprehensive resistance identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110082298A CN110082298A (en) | 2019-08-02 |
| CN110082298B true CN110082298B (en) | 2020-05-19 |
Family
ID=67420191
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910401015.1A Active CN110082298B (en) | 2019-05-15 | 2019-05-15 | Hyperspectral image-based wheat variety gibberellic disease comprehensive resistance identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110082298B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110763698B (en) * | 2019-10-12 | 2022-01-14 | 仲恺农业工程学院 | Hyperspectral citrus leaf disease identification method based on characteristic wavelength |
| CN110736750B (en) * | 2019-10-28 | 2022-03-04 | 安徽大学 | Wheat scab detection method based on multi-angle field high-definition imaging |
| CN110596280B (en) * | 2019-10-29 | 2022-04-05 | 南京财经大学 | Rapid detection method for wheat vomitoxin pollution level based on hyperspectral image and spectrum information fusion |
| CN112880734A (en) * | 2020-12-31 | 2021-06-01 | 中农新科(苏州)有机循环研究院有限公司 | Biological drying process digital monitoring system for reactor |
| CN116523866B (en) * | 2023-04-26 | 2023-12-01 | 扬州大学 | Wheat scab resistance identification method, system, electronic equipment and storage medium |
| CN116973317A (en) * | 2023-06-13 | 2023-10-31 | 河北农业大学 | Method for identifying resistance of plant to plutella xylostella based on hyperspectral imaging technology |
| CN116821698B (en) * | 2023-08-31 | 2024-01-05 | 中国科学技术大学 | Wheat scab spore detection method based on semi-supervised learning |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102072885A (en) * | 2010-12-06 | 2011-05-25 | 浙江大学 | Machine vision-based paddy neck blast infection degree grading method |
| CN103091284A (en) * | 2013-02-28 | 2013-05-08 | 河南工业大学 | Near-infrared spectrum technology based method for rapidly identifying wheat grains with FHB (Fusarium Head Blight) |
| CN106777845A (en) * | 2017-03-22 | 2017-05-31 | 南京农业大学 | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method |
| CN107527326A (en) * | 2017-08-17 | 2017-12-29 | 安徽农业大学 | A kind of wheat scab diagnostic method based on high light spectrum image-forming |
| CN107576618A (en) * | 2017-07-20 | 2018-01-12 | 华南理工大学 | Pyricularia Oryzae detection method and system based on depth convolutional neural networks |
| CN109035231A (en) * | 2018-07-20 | 2018-12-18 | 安徽农业大学 | A kind of detection method and its system of the wheat scab based on deep-cycle |
| CN109187552A (en) * | 2018-08-30 | 2019-01-11 | 安徽农业大学 | A kind of gibberella saubinetii disease grade determination method based on cloud model |
| CN109544538A (en) * | 2018-11-27 | 2019-03-29 | 安徽大学 | Wheat scab disease grade grading method and device |
| CN109657653A (en) * | 2019-01-21 | 2019-04-19 | 安徽大学 | A kind of wheat seed head blight recognition methods based on Imaging Hyperspectral Data |
-
2019
- 2019-05-15 CN CN201910401015.1A patent/CN110082298B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102072885A (en) * | 2010-12-06 | 2011-05-25 | 浙江大学 | Machine vision-based paddy neck blast infection degree grading method |
| CN103091284A (en) * | 2013-02-28 | 2013-05-08 | 河南工业大学 | Near-infrared spectrum technology based method for rapidly identifying wheat grains with FHB (Fusarium Head Blight) |
| CN103091284B (en) * | 2013-02-28 | 2015-07-15 | 河南工业大学 | Near-infrared spectrum technology based method for rapidly identifying wheat grains with FHB (Fusarium Head Blight) |
| CN106777845A (en) * | 2017-03-22 | 2017-05-31 | 南京农业大学 | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method |
| CN107576618A (en) * | 2017-07-20 | 2018-01-12 | 华南理工大学 | Pyricularia Oryzae detection method and system based on depth convolutional neural networks |
| CN107527326A (en) * | 2017-08-17 | 2017-12-29 | 安徽农业大学 | A kind of wheat scab diagnostic method based on high light spectrum image-forming |
| CN109035231A (en) * | 2018-07-20 | 2018-12-18 | 安徽农业大学 | A kind of detection method and its system of the wheat scab based on deep-cycle |
| CN109187552A (en) * | 2018-08-30 | 2019-01-11 | 安徽农业大学 | A kind of gibberella saubinetii disease grade determination method based on cloud model |
| CN109544538A (en) * | 2018-11-27 | 2019-03-29 | 安徽大学 | Wheat scab disease grade grading method and device |
| CN109657653A (en) * | 2019-01-21 | 2019-04-19 | 安徽大学 | A kind of wheat seed head blight recognition methods based on Imaging Hyperspectral Data |
Non-Patent Citations (6)
| Title |
|---|
| Fusarium Toxins in Chinese Wheat since the 1980s;Jianbo Qiu et al;《MDPI》;20190430;全文 * |
| Hyperspectral imaging for detection of scab in wheat;Stephen R. Delwiche et al;《SPIE》;20001229;全文 * |
| 基于动态时间弯曲的马铃薯干时序高光谱诊断方法腐病发病期;金秀等;《食品科学》;20181231;第39卷(第19期);全文 * |
| 基于高光谱成像技术的小麦籽粒赤霉病识别;梁琨等;《农业机械学报》;20160228;第47卷(第2期);全文 * |
| 小麦抗虫性评价技术规范第4部分：小麦抗赤霉病评价技术规范;中华人民共和国农业部;《中华人民共和国农业行业标准NY/T 1443.4-2007》;20071201;全文 * |
| 松材线虫病害高光谱时序与敏感特征研究;黄明祥等;《遥感技术与应用》;20121231;第27卷(第6期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110082298A (en) | 2019-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110082298B (en) | Hyperspectral image-based wheat variety gibberellic disease comprehensive resistance identification method | |
| CN101881726B (en) | Nondestructive detection method for comprehensive character living bodies of plant seedlings | |
| Li et al. | Machine vision technology for detecting the external defects of fruits—a review | |
| CN105158186B (en) | A kind of method detected based on high spectrum image to ternip evil mind | |
| CN107392920B (en) | Plant health distinguishing method and device based on visible light-terahertz light | |
| CN104990892B (en) | The spectrum picture Undamaged determination method for establishing model and seeds idenmtification method of seed | |
| CN105067532B (en) | A kind of method for differentiating sclerotinia sclerotiorum and gray mold early stage scab | |
| Zou et al. | Peanut maturity classification using hyperspectral imagery | |
| Ji et al. | In-field automatic detection of maize tassels using computer vision | |
| CN107679569A (en) | Raman spectrum substance automatic identifying method based on adaptive hypergraph algorithm | |
| CN111199192A (en) | Method for detecting integral maturity of field red globe grapes by adopting parallel line sampling | |
| Archana et al. | A novel method to improve computational and classification performance of rice plant disease identification | |
| CN103528967A (en) | Hyperspectral image based overripe Lonicera edulis fruit identification method | |
| CN111562235A (en) | Method for rapidly identifying black-leaf outbreak disease and infection degree of tobacco leaves based on near infrared spectrum | |
| CN114332534B (en) | Hyperspectral image small sample classification method | |
| CN106568730B (en) | A kind of rice yin-yang leaf fringe recognition methods based on Hyperspectral imaging near the ground | |
| Zhang et al. | Chlorophyll content detection of field maize using RGB-NIR camera | |
| Gaikwad et al. | Multi-spectral imaging for fruits and vegetables | |
| CN104297136A (en) | Hyperspectral image-based method for forecasting growth of pseudomonas aeruginosa | |
| CN113947796A (en) | Human body temperature trend detection method and device based on identity recognition | |
| CN109164069A (en) | A kind of identification method of fruit tree foliage disease rank | |
| CN112528726A (en) | Aphis gossypii insect pest monitoring method and system based on spectral imaging and deep learning | |
| Xu et al. | A Band Math-ROC operation for early differentiation between sclerotinia sclerotiorum and botrytis cinerea in oilseed rape | |
| Benlachmi et al. | Fruits Disease Classification using Machine Learning Techniques | |
| Xu et al. | Early identification of Curvularia lunata based on hyperspectral imaging |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 210031 No. 40, Taiwan Road, Pukou District, Nanjing, Jiangsu Applicant after: NANJING AGRICULTURAL University Address before: 211225, Jiangsu, Nanjing province Lishui Baima Town National Agricultural Science and Technology Park, Nanjing Agricultural University base Applicant before: NANJING AGRICULTURAL University |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |