CN107748165A - Potato solanine detection method based on machine vision Yu electronic nose integration technology - Google Patents

Potato solanine detection method based on machine vision Yu electronic nose integration technology Download PDF

Info

Publication number
CN107748165A
CN107748165A CN201711127137.3A CN201711127137A CN107748165A CN 107748165 A CN107748165 A CN 107748165A CN 201711127137 A CN201711127137 A CN 201711127137A CN 107748165 A CN107748165 A CN 107748165A
Authority
CN
China
Prior art keywords
potato
electronic nose
solanine
sample
machine vision
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.)
Pending
Application number
CN201711127137.3A
Other languages
Chinese (zh)
Inventor
黄星奕
孙兆燕
任晓锋
田潇瑜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu University
Original Assignee
Jiangsu University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu University filed Critical Jiangsu University
Priority to CN201711127137.3A priority Critical patent/CN107748165A/en
Publication of CN107748165A publication Critical patent/CN107748165A/en
Pending legal-status Critical Current

Links

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
    • G01N21/84—Systems specially adapted for particular applications
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N30/00—Investigating or analysing materials by separation into components using adsorption, absorption or similar phenomena or using ion-exchange, e.g. chromatography or field flow fractionation
    • G01N30/02—Column chromatography
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N30/00—Investigating or analysing materials by separation into components using adsorption, absorption or similar phenomena or using ion-exchange, e.g. chromatography or field flow fractionation
    • G01N30/02—Column chromatography
    • G01N30/04—Preparation or injection of sample to be analysed
    • G01N30/06—Preparation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004—Gaseous mixtures, e.g. polluted air
    • G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
    • G01N33/0027—General constructional details of gas analysers, e.g. portable test equipment concerning the detector
    • G01N33/0031—General constructional details of gas analysers, e.g. portable test equipment concerning the detector comprising two or more sensors, e.g. a sensor array
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • 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
    • G01N21/84—Systems specially adapted for particular applications
    • G01N2021/8466—Investigation of vegetal material, e.g. leaves, plants, fruits
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N30/00—Investigating or analysing materials by separation into components using adsorption, absorption or similar phenomena or using ion-exchange, e.g. chromatography or field flow fractionation
    • G01N30/02—Column chromatography
    • G01N30/04—Preparation or injection of sample to be analysed
    • G01N30/06—Preparation
    • G01N2030/062—Preparation extracting sample from raw material

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Biochemistry (AREA)
  • Analytical Chemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Engineering & Computer Science (AREA)
  • Medicinal Chemistry (AREA)
  • Food Science & Technology (AREA)
  • Combustion & Propulsion (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

The invention discloses the potato solanine detection method based on machine vision Yu electronic nose integration technology, belong to nondestructive measuring method of the farm product technical field.Potato image is obtained with machine vision, extracts the color and textural characteristics in shagreen and region of germinateing in image；Then the electronic nose sensor array of potato measure is exclusively used according to gas information architecture obtained by gas chromatography-mass spectrometry, optimize gas collecting apparatus and condition determination, gas componant collection is carried out to potato to be measured, extracts the stationary value of response signal as characteristic value；Most gained image information carries out Feature-level fusion with electronic nose information at last, establishes its correlation model between black nightshade cellulose content, realizes the rapidly and efficiently lossless decomposition to Analysis of Solanine in Potato content.The present invention proposes to be predicted black nightshade cellulose content during storage of potato with machine vision integration technology based on electronic nose, has not yet to see report, is that non-destructive testing technology is applied into trying to explore for agricultural product security Indexs measure.

Description

Potato solanine detection method based on machine vision Yu electronic nose integration technology
Technical field
The present invention relates to the potato solanine detection method based on machine vision Yu electronic nose integration technology, belong to agricultural production Product technical field of nondestructive testing.
Background technology
Potato is rich in multiple nutritional components such as carbohydrate, protein, mineral matter (phosphorus, calcium etc.) and vitamins. It is the fourth-largest cereal crops in the world, while is also vegetables and industrial crops important in the world, cultivation range throughout is global. At present, China's potato planting area and total output occupy first place in the world.With the expansion in potato yield and consumption face, Ma Ling The edible safety problem of potato and its product is also more and more concerned.In potato storage and transportation in shagreen and germination individual Solanine (main component is α-solanine and α-chaconine) content increases severely, and solanine has stronger toxicity, human body can be disappeared Change system, nervous system and gastronintestinal system are caused to damage, and death can be caused when serious, therefore to the inspection of Analysis of Solanine in Potato Survey becomes particularly important.
Detection to shagreen and the sample that germinates at this stage is broadly divided into conventional method, chemical method detection and Non-Destructive Testing Method.Conventional method wastes time and energy by manually being rejected to greening and germination individual；Conventional chemical detection method is divided into Ultraviolet spectrophotometry, high-efficient liquid phase technique etc., sample is not only damaged, is wasted time and energy and higher are required to operating personnel.At this stage Conventional lossless detection method is mainly the detection such as near infrared detection, EO-1 hyperion detection, machine vision, through retrieving Patents Have：A kind of greening potato detection stage division based on machine vision, Publication No.：CN103394472A.The Patent exploitation A kind of method for carrying out detecting classification to the potato of epidermis greening based on tone value diagnostic method.The patent does not stop to roll to each The image of the dynamic width different surfaces of sample collection three, extracts R, G, B value of each image and is converted into tone value, by with threshold value ratio After count its greening point and normally count out, most sample is divided into normal, unobvious greening and afforests sample at last, passes through reality Its accuracy rate is up to 90.8% after example checking.Although this detection method improves detection speed compared with artificial detection, to Ma Ling Potato has carried out further division, but this method of discrimination divides only in accordance with potato exterior color feature to it, to it Internal black nightshade cellulose content is not measured, not can determine that whether sample is edible.Therefore explore a kind of using black nightshade cellulose content to sentence Determining the fast non-destructive detection method of foundation becomes particularly important.
Exterior color information when this apllied patent is not only in accordance with potato greening and germination, and it is differentiated simultaneously The change of internal odiferous information, it is combined using machine vision with Electronic Nose Technology and potato inside black nightshade cellulose content is determined Amount analysis, potato is divided into edible potato, green potato and germination according to black nightshade cellulose content and potato surface color feature Potato, realize effectively classification.The method can not only ensure the edible safety of potato, and green sample can be used for medical science Research, and the sample that germinates can then be used as potato seed, fully achieve the recycling of resource.By machine vision and electronic nose skill in this patent Art, which is combined, can make up single detection technique to the defects of initial stage, sample identification rate was low.This invention uses lossless detection method Using black nightshade cellulose content as according to being divided to sample, the edible safety of only potato and its product does not provide guarantor Barrier, the development of potato process deeply industry is may additionally facilitate, China's future Development of Potato Industry is had a far reaching influence.
The content of the invention
In order to overcome problem present in background technology, it is an object of the invention to provide one kind based on machine vision and electricity The potato solanine detection method of sub- nose integration technology.Analysis of Solanine in Potato content of this patent according to efficient liquid phase measure Potato is divided into edible and not edible；Potato image information is obtained using computer vision, image is pre-processed, Remove background, shagreen and germination region segmentation, extraction color feature value；Potato odiferous information, data are gathered by electronic nose Its stationary value is extracted after pretreatment as characteristic value；The image information of gained is melted with electronic nose information in characteristic layer Close, and the correlation model established between its black nightshade cellulose content obtained by efficient liquid phase, most potato is divided into edible potato, green at last Skin potato and germination potato.The image information of potato is combined with odiferous information, to potato on the premise of sample is not damaged Middle black nightshade cellulose content is predicted, so as to fast and efficiently realize the whether edible qualitative discrimination of potato.
The technical solution adopted by the present invention is as follows:Potato solanine based on machine vision and electronic nose integration technology is examined Survey method, carry out as steps described below：
Potato image is obtained using first (Charge Coupled Device, the CCD) camera of Charged Couple first, it is right Image is pre-processed, and extracts the color and textural characteristics in its shagreen and region of germinateing；Then according to gas chromatography-mass spectrography Gas componant information architecture obtained by instrument (Gas Chromatography-Mass Spectrometer, GC-MS) is exclusively used in Ma Ling The electronic nose sensor array of potato measure, selected sample gas collecting apparatus, optimizes condition determination, and the potato of different qualities is carried out Gas componant is gathered, and the data obtained is pre-processed, extracts its stationary value as characteristic value；Most at last gained image information Feature-level fusion is carried out with electronic nose information, establishes its correlation model between black nightshade cellulose content, is realized to imperial in potato The rapidly and efficiently lossless decomposition of certain herbaceous plants with big flowers cellulose content.
Specific implementation step：Potato solanine detection method based on machine vision Yu electronic nose integration technology, according to Following step is carried out：
Step 1：Using the potato image capturing system specially designed in machine vision, image is carried out to potato sample Collection, it is ensured that the integrality of potato surface color and texture feature information, gained image is preserved into computer.
Step 2：Headspace gas sampling, recorded electronic nose sensing are carried out to potato sample using the electronic nose specially designed Device array response signal value, so as to obtain response curve of the sensor array to different detection samples, it is stored in computer In.
Step 3：To potato Sample pretreatment, it is measured, is obtained using high performance liquid chromatography after extracting its solanine Obtain its solanine content value.
Step 4：Potato sample foreign information to the reflection of step 1 machine vision acquired image and step 2 electricity respectively Potato sample interior information is pre-processed measured by sub- nose, and extracts respective characteristic value, and the two is melted in characteristic layer Close, then build its dependency relation between black nightshade cellulose content, establish model and potato shagreen and germination state are sentenced Not.
The potato image capturing system wherein specially designed in machine vision in step 1 by light box, light source, objective table, Image first-class several major composition.Light box is that top is hemispherical, a diameter of designed by the needs shot according to potato 38cm；Bottom is straight barrel type, diameter 38cm, high 15cm.Light source uses adjustable ring-shaped light emitting diode (Light Emitting Diode, LED), the distance of light source and sample can adjust according to the difference of Potato Cultivars and size.CCD camera shooting image When should adjust light source position and intensity so that it is radiated at uniform light and moderate strength on sample, without obvious flare.
Electronic nose wherein in step 2 specially designs, by Jiangsu University's Non-Destructive Testing team itself development.The electronics Nose is made up of gas sampling system, gas sensor array, master control system and the part of software analysis system four.Wherein, air-sensitive passes The determination process of sensor array is：Gas componant caused by potato sample is detected using GC-MS, according to GC-MS Acquired results, it is applied to the sensor array of potato solanine detection with reference to the interaction sensitivity characteristic structure of electronic nose sensor Row.The analysis result that data are surveyed to GC-MS shows that potato gas componant mainly includes 30 kinds, wherein 9 kinds of alcohols, aldehydes 8 Kind, 6 kinds of hydro carbons also include the materials such as ketone, acids, esters, furans.It is final to determine that sensor array is classified as：TGS832、 WSP2110、TGS822、TGS813、TGS880、TGS2610、TGS2611、TGS2600、TGS2620、MQ135、MQ137、 MQ316.Gas collecting apparatus, gas collection time, sample collection time are optimized before sample collection wherein in step 2, through experiment Gas collecting apparatus used is the tasteless glass case of 800ml cylinders after optimization, and the gas collection time is 40min, and the sample collection time is 420s.
The high performance liquid chromatography of solanine assay described in step 3, is carried out as steps described below：By sample not Material ratio 1 is pressed after removing the peel and directly crushing:40 add the acetic acid that volumetric concentration is 5%, at room temperature in ultrasonic 20min extractions sample Solanine；Filter, the acetic acid resuspension that filter residue volumetric concentration is 5% filters 2 times, and all filtrates are merged into a conical flask It is interior, add concentrated ammonia liquor regulation pH to 10-11 and solanine precipitates.Under the conditions of 70 DEG C of alkaline solution 4 are placed in after water-bath 50min DEG C refrigerator overnight.Alkaline solution 18000r/min is centrifuged into 30min, centrifugation is cleaned multiple times with the ammoniacal liquor that volumetric concentration is 2% Clarified to solution.60 DEG C of forced air dryings of gained sediment, are then completely dissolved in tetrahydrofuran/acetonitrile/20mmol KH2PO4(5:3: 2) in solution and centrifuge, take 2ml supernatants to be used for efficient liquid phase chromatographic analysis.
The extraction process for the potato sample foreign information characteristics value that machine vision described in step 4 obtains, under State step progress：Extract image in potato color and texture information, by RGB image color value be separately converted to HSV and YCbCr tone values；(H values) split plot design is tieed up using tone and obtains complete bianry image.The shagreen in image and germination region are extracted, Green point color value N in region is extracted using point by point scanning method, it is right to extract each color value institute using Marius region-filling algorithms The pixel number S answered, the corresponding relative ratio α of its R, G, B, H, Ycb, Cr value is obtained respectively, such asWherein NjFor a certain H values corresponding to full sample, SjFor this, H values are right The pixel number answered；niThe a certain H values corresponding to green and germination region for extraction；siCorresponding pixel of H values for this Number.Its required each ratio is extracted image feature value.
The extraction process for the potato sample interior information characteristics value that electronic nose described in step 4 obtains, according to following Step is carried out：The stationary value on sensor array response curve obtained by extraction step 2 is as electronic nose characteristic value.Using Principal component analysis carries out dimension-reduction treatment to all characteristic value datas of gained.
Discrimination model establishes process in step 4, carries out as steps described below：Electronic nose information after back dimensionality reduction is special Value indicative is merged with the external information characteristic value that machine vision obtains using fuzzy set theory in characteristic layer, and fuse information is made For the input of model, using solanine content value obtained by efficient liquid phase as output trained values, pass through RBF (Radial Basis Function, RBF) neutral net and independent variable method build non-linear decision model, to Analysis of Solanine in Potato Content, which is made, to be appropriately determined, and then sample is divided into edible potato, shagreen potato and germination potato.
The invention has the advantages that:The present invention is proposed based on electronic nose with machine vision integration technology to potato Black nightshade cellulose content is predicted in storage, has not yet to see report, is that non-destructive testing technology is applied into agricultural product security Property Indexs measure is tried to explore.
Brief description of the drawings
Fig. 1 is machine vision image collecting device, wherein 1 computer, 2CCD cameras, 3 be circular lamp, 4 objective tables, 5 IMAQ case..
Fig. 2 is the electronic nose gathered data of sample, and abscissa is the sampling time, and ordinate is sensor response.
Embodiment
The present invention is described in more detail with reference to embodiments.
Example：
(1) collection of test sample：The intact and shagreen potato samples totally 96 that selection peasant has just harvested, germinate sample 20, put it into climatic chamber and store.Normal and micro- 9, green sample, special green 4, sample, germination sample were chosen every 4 days This 3 totally 16 samples, clean remained on surface soil stain, are tested after its dry tack free.
(2) IMAQ：In the IMAQ case (5 in figure) that sample is positioned over shown in Fig. 1 on objective table (4 in figure), Wherein the distance between sample and CCD camera (2 in figure) are adjusted according to the difference of Potato Cultivars, for this batch sample This its optimum distance is 6.5cm；The position of circular lamp (3 in figure) is adjusted during image taking, when circular lamp with sample at a distance of 5.2cm When be radiated at uniform light and moderate strength on sample, without obvious flare.Each potato samples shoot two images, Image and preservation when respectively positive and negative two sides is placed is into 1 computer.
(3) background segment, filtering, rim detection are carried out to the potato samples image collected in step (2).Extraction The rgb value of potato in image, is separately converted to HSV and YCbCr values；(H) split plot design is tieed up using tone and obtains complete binary map As and extract the area-of-interest in image herein for shagreen and germination region.Green point in region is extracted using point by point scanning method Color value N, the pixel number S corresponding to each color value is extracted by Marius region-filling algorithms, obtain respectively its R, G, B, the corresponding relative ratio α of H, Ycb, Cr value, the relative ratio of gained can be used as the characteristic value of potato image.Through it is main into Preceding 3 principal components are selected for the foundation of later stage decision model and the prediction of unknown sample black nightshade cellulose content after dividing dimensionality reduction.
(4) electronic nose detection process is as follows：Detected using Non-Destructive Testing team of Jiangsu University self-control electronic nose.Will choosing The potato samples taken are put into the tasteless glass container of 800ml cylinders and carry out gas collection, and the gas collection time is 40min, so as to obtain Headspace gas；Electronic nose preheats 3h, before each electronic nose detection starts, electric nasus system is reduced using oxygen, reduced After the completion of, electronic nose extracts the headspace gas in closed container, and potato samples are detected；In detection process according to 1 time/ The speed record sensor response signal of second, gathers 420 signal values, so as to obtain sensor array to different detection samples altogether Response curve.
(5) determination process of electronic nose gas sensor array is in step (4)：Using GC-MS to potato sample institute Caused gas componant is detected, and according to GC-MS acquired results, the interaction sensitivity characteristic structure with reference to electronic nose sensor is suitable Sensor array for the detection of potato solanine.The analysis result that data are surveyed to GC-MS shows, potato gas componant It is main to include 30 kinds, analyzed from high to low according to its gas percentage composition, wherein 9 kinds of alcohols：2- methyl n-butanol, 1- are pungent Dilute -3- alcohol, n-amyl alcohol, phenmethylol, 3,7- dimethyl -2,6- octadiene -1- alcohol, 2- ethyls-n-hexyl alcohol, 1-POL, 3- Dilute -1- the alcohol of methyl n-butanol, 2- penta；8 kinds of aldehydes：2- octenals, 2- methyl n-butanal, 2- methyl-propionic aldehyde, phenylacetaldehyde, 3- first Base n-butanal, benzaldehyde, hexanal, the olefine aldehydrs of 2,4- bis-；6 kinds of hydro carbons：Hexane, 1- methyl -2- ethyl cyclopentanes, pentane, 3- ethyls - 2- methyl isophthalic acids, oneself two dilute, d-limonens of 3-；Also include the materials such as ketone, acids, lipid, furans.It is final to determine sensor Array is：TGS832、WSP2110、TGS822、TGS813、TGS880、TGS2610、TGS2611、TGS2600、TGS2620、 MQ135、MQ137、MQ316。
(6) electronic nose sensor array response signal to the sensor array response signal obtained as shown in Fig. 2 carry out Feature extraction.According to research before, the value of plateau has optimal differentiation performance in sensor response curve, therefore selects The average value of 380s-400s stable responses value in each sensor response curve is selected as characteristic value.Using PCA Dimensionality reduction is carried out to gained characteristic value, finally selects foundation and unknown sample black nightshade that preceding 3 principal components are used for later stage decision model The prediction of cellulose content.
(7) efficient liquid phase detection process is as follows：Solanine extraction is extracted using the heavy principle of the molten alkali of acid, concrete operations For：Material 1g is taken to add the acetic acid that 40ml volumetric concentrations are 5% after sample is directly crushed without peeling, it is ultrasonic at room temperature Solanine in 20min extraction samples；Filter, the acetic acid resuspension that filter residue volumetric concentration is 5% filters 2 times, all filtrates It is merged into a conical flask, adds concentrated ammonia liquor regulation pH to 10-11 and solanine precipitates.70 DEG C of conditions of alkaline solution 4 DEG C of refrigerator overnights are placed in after lower water-bath 50min.Alkaline solution 18000r/min is centrifuged into 30min, is 2% with volumetric concentration Ammoniacal liquor cleans 3 centrifugations to solution and clarified.60 DEG C of forced air dryings of gained sediment, be then completely dissolved in tetrahydrofuran/acetonitrile/ 20mmolKH2PO4In solution and centrifuge, take 2ml supernatants to be analyzed for HPLC.Its HPLC condition is chromatographic column, Inersil NH2(5um, 4.0mm × 250mm)；Mobile phase, acetonitrile/20mmol KH2PO4(80:20, v/v)；Flow velocity, 1.0ml/min；Column temperature For room temperature；Detection wavelength 210nm；Applied sample amount 20ul.Liquid phase acquired results are analyzed, draw the specific of solanine in sample Content.
(8) 3 external information features for obtaining 3 internal information characteristic values of the electronic nose after dimensionality reduction and machine vision Value carries out Feature-level fusion using fuzzy set theory, and A is regarded as to the set of the possible decision-making of system, B is regarded as to the set of sensor, A and B relational matrix RA*BIn element uijRepresent that sensor i infers the possibility that decision-making is j, X represents what each sensor judged Confidence level, the Y obtained by blurring mapping are exactly the possibility of each decision-making.Specifically, 6 sensors enter system in this example Row observation, and the possible decision-making of system has 3, then：
A={ y1/ decision-making 1, y2/ decision-making 2, y3/ decision-making 3 }；
B={ x1/ sensor 1, x2/ sensor 2 ..., x6/ sensor 6 }
Judgement of the sensor to each possible decision-making is represented with the membership function being defined on A.If sensor is sentenced to system Disconnected result is：
[ui1/ decision-making 1, ui2/ decision-making 2, ui3/ decision-making n] 0≤uij≤1
Think the possibility u that result is decision-making jij, it is denoted as vector (ui1, ui2..., uin), then 6 sensors form 3*6 Relational matrix be：
The confidence level of each sensor is subordinate to several X={ x with B1/ sensor 1, x2/ sensing Device 2 ..., x6/ sensor 6 } represent, then according to Y=XRA*BCarry out blurring mapping, so that it may draw Y=(y1,y2,y3), i.e., The possibility y of each decision-making after comprehensive descisioni。
(9) for fuse information as the input of model, solanine content value obtained by efficient liquid phase, which is used as, exports trained values, passes through RBF neural and independent variable method build non-linear decision system, and Analysis of Solanine in Potato content is judged, and then Sample is divided into edible potato, shagreen potato and germination potato.RBF neural is by three layers i.e. input layer, hidden layer and output layer group Into.Input layer functions only as transmitting the effect of signal；Hidden layer is that the parameter of activation primitive is adjusted, using non-thread Property optimisation strategy.The parameter that RBF neural learning algorithm needs to solve has 3：Center, variance and the hidden layer of basic function To the weights of output layer.This example uses the RBF neural learning method of Self-organizing Selection Center, and uses Gaussian function conduct RBF.It is specific to represent as follows:
In formula, | | xp-ci| | it is European norm；C is the center of Gaussian function；σ is the variance of Gaussian function.
The output of network is
In formula,For p-th of input sample；P=1,2,3 ..., 96；ciFor network hidden layer The center of node；wiFor the connection weight of hidden layer to output layer；I=1,2,3 ..., h.h are node in hidden layer；yiFor with it is defeated Enter the reality output of j-th of output node of network corresponding to sample.
If d is the desired output of sample, then the variance of basic function is represented by
Experiment the data obtained is handled by the RBF neural of structure, Analysis of Solanine in Potato content is made It is appropriately determined, and then sample is divided into edible potato, shagreen potato and germination potato.Its result of determination is：Sentence when number of principal components is 6 It is optimal to determine effect, wherein training set accuracy rate is 96.15%, and test set accuracy rate is 90.84%, sample has been carried out well Division；The potato solanine proposed by the invention based on machine vision and electronic nose integration technology is also demonstrated simultaneously to detect Method is to potato shagreen and the feasibility for pattern detection of germinateing.Finally constructed system has been carried out using large sample experiment It is kind, by realizing the kind judging of sample to the prediction of Analysis of Solanine in Potato content.

Claims (7)

1. the potato solanine detection method based on machine vision Yu electronic nose integration technology, it is characterised in that according to following steps It is rapid to carry out：
Step 1：Using the potato image capturing system specially designed in machine vision, image is carried out to potato sample and adopted Collection, it is ensured that the integrality of potato surface color and texture feature information, gained image is preserved into computer；
Step 2：Headspace gas sampling, recorded electronic nose sensor array are carried out to potato sample using the electronic nose specially designed Row response signal value, so as to obtain response curve of the sensor array to different detection samples, it is stored in computer；
Step 3：To potato Sample pretreatment, it is measured after extracting its solanine using high performance liquid chromatography, obtains it Solanine content value；
Step 4：Respectively to the potato sample foreign information and step 2 electronic nose of the reflection of step 1 machine vision acquired image Measured potato sample interior information is pre-processed, and extracts respective characteristic value, the two is merged in characteristic layer, so Its dependency relation between black nightshade cellulose content is built afterwards, establishes model and potato shagreen and germination state are differentiated.
2. the potato solanine detection method according to claim 1 based on machine vision Yu electronic nose integration technology, It is characterized in that
The potato image capturing system wherein specially designed in machine vision in step 1 is by light box, light source, objective table, shooting First-class several major composition；Light box is that top is hemispherical, a diameter of 38cm designed by the needs shot according to potato； Bottom is straight barrel type, diameter 38cm, high 15cm；Light source uses adjustable ring-shaped light emitting diode (LED), light source and sample away from It can adjust from according to the difference of Potato Cultivars and size；Light source position and intensity should be adjusted during CCD camera shooting image to be made Its uniform light and moderate strength for being radiated on sample is obtained, without obvious flare.
3. the potato solanine detection method according to claim 1 based on machine vision Yu electronic nose integration technology, It is characterized in that
Electronic nose wherein in step 2 specially designs, by Jiangsu University's Non-Destructive Testing team itself development；The electronic nose by Gas sampling system, gas sensor array, master control system and the part of software analysis system four composition；Wherein, gas sensor The determination process of array is：Gas componant caused by potato sample is detected using GC-MS, according to obtained by GC-MS As a result, the interaction sensitivity characteristic structure with reference to electronic nose sensor is applied to the sensor array that potato solanine detects；It is right The analysis result that GC-MS surveys data shows that potato gas componant mainly includes 30 kinds, wherein 9 kinds of alcohols, 8 kinds of aldehydes, hydrocarbon 6 kinds of class also includes the materials such as ketone, acids, esters, furans；It is final to determine that sensor array is classified as：TGS832、WSP2110、 TGS822、TGS813、TGS880、TGS2610、TGS2611、TGS2600、TGS2620、MQ135、MQ137、MQ316；Wherein walk Gas collecting apparatus, gas collection time, sample collection time are optimized before sample collection in rapid 2, the collection used after optimum experimental Device of air is the tasteless glass case of 800ml cylinders, and the gas collection time is 40min, and the sample collection time is 420s.
4. the potato solanine detection method according to claim 1 based on machine vision Yu electronic nose integration technology, It is characterized in that
The high performance liquid chromatography of solanine assay described in step 3, is carried out as steps described below：By sample without going Skin presses material ratio 1 after directly crushing:40 add the acetic acid that volumetric concentration is 5%, at room temperature the black nightshade in ultrasonic 20min extractions sample Element；Filtering, the acetic acid resuspension that filter residue volumetric concentration is 5% is filtered 2 times, and all filtrates are merged into a conical flask, Add concentrated ammonia liquor regulation pH to 10-11 solanine precipitates；Under the conditions of 70 DEG C of alkaline solution 4 DEG C are placed in after water-bath 50min Refrigerator overnight；Alkaline solution 18000r/min is centrifuged into 30min, centrifugation is cleaned multiple times extremely with the ammoniacal liquor that volumetric concentration is 2% Solution is clarified；60 DEG C of forced air dryings of gained sediment, are then completely dissolved in tetrahydrofuran/acetonitrile/20mmol KH2PO4(5:3:2) In solution and centrifuge, take 2ml supernatants to be used for efficient liquid phase chromatographic analysis.
5. the potato solanine detection method according to claim 1 based on machine vision Yu electronic nose integration technology, It is characterized in that
The extraction process for the potato sample foreign information characteristics value that machine vision described in step 4 obtains, according to following steps It is rapid to carry out：The color and texture information of potato in image are extracted, RGB image color value is separately converted to HSV and YCbCr colors Tone pitch；(H values) split plot design is tieed up using tone and obtains complete bianry image；Extract image in shagreen and germination region, using by Spot scan method extracts region Green point color value N, and the picture corresponding to each color value is extracted using Marius region-filling algorithms Vegetarian refreshments number S, the corresponding relative ratio α of its R, G, B, H, Ycb, Cr value is obtained respectively, such as Wherein NjFor a certain H values corresponding to full sample, SjCorresponding pixel number of H values for this；niTo carry A certain H values corresponding to the green taken and germination region；siCorresponding pixel number of H values for this；Its required each ratio is For the image feature value extracted.
6. the potato solanine detection method according to claim 1 based on machine vision Yu electronic nose integration technology, It is characterized in that
The extraction process for the potato sample interior information characteristics value that electronic nose described in step 4 obtains, as steps described below Carry out：The stationary value on sensor array response curve obtained by extraction step 2 is as electronic nose characteristic value；Using it is main into Analysis carries out dimension-reduction treatment to all characteristic value datas of gained.
7. the potato solanine detection method according to claim 1 based on machine vision Yu electronic nose integration technology, It is characterized in that
Discrimination model establishes process in step 4, carries out as steps described below：By the electronic nose information characteristics value after back dimensionality reduction The external information characteristic value obtained with machine vision is merged using fuzzy set theory in characteristic layer, using fuse information as mould The input of type, using solanine content value obtained by efficient liquid phase as output trained values, pass through RBF (RBF) neutral net Non-linear decision model is built with independent variable method, Analysis of Solanine in Potato content is made and is appropriately determined, and then by sample It is divided into edible potato, shagreen potato and germination potato.
CN201711127137.3A 2017-11-15 2017-11-15 Potato solanine detection method based on machine vision Yu electronic nose integration technology Pending CN107748165A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711127137.3A CN107748165A (en) 2017-11-15 2017-11-15 Potato solanine detection method based on machine vision Yu electronic nose integration technology

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711127137.3A CN107748165A (en) 2017-11-15 2017-11-15 Potato solanine detection method based on machine vision Yu electronic nose integration technology

Publications (1)

Publication Number Publication Date
CN107748165A true CN107748165A (en) 2018-03-02

Family

ID=61251983

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711127137.3A Pending CN107748165A (en) 2017-11-15 2017-11-15 Potato solanine detection method based on machine vision Yu electronic nose integration technology

Country Status (1)

Country Link
CN (1) CN107748165A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108827889A (en) * 2018-06-13 2018-11-16 江西中医药大学 A kind of glue class material discrimination method based on optical characteristics
CN109307638A (en) * 2018-08-07 2019-02-05 江苏大学 A kind of measuring method and device of steamed bun specific volume
CN110133049A (en) * 2019-04-10 2019-08-16 浙江大学 Tea grades fast non-destructive detection method based on electronic nose and machine vision
CN110702768A (en) * 2019-11-05 2020-01-17 广东省农业科学院农产品公共监测中心 Agricultural product quality production line dynamic lossless bionic olfaction detection equipment and method
CN111443160A (en) * 2020-01-23 2020-07-24 华东理工大学 Gas-sensitive-gas chromatography information fusion and electronic nose instrument on-line analysis method
CN112036470A (en) * 2020-08-28 2020-12-04 扬州大学 Cloud transmission-based multi-sensor fusion cucumber bemisia tabaci identification method
WO2021147274A1 (en) * 2020-01-23 2021-07-29 华东理工大学 Gas-sensing-gas-phase electronic nose instrument and fermentation-malodor multi-state parameter online analysis method
CN115578553A (en) * 2022-11-22 2023-01-06 河南知微生物工程有限公司 Method for quickly detecting formaldehyde based on time sequence image sequence

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
徐海霞: "《基于机器视觉和电子鼻技术的菠菜新鲜度无损检测研究》", 《中国优秀硕士学位论文全文数据库（电子期刊）》 *
王守兰 等: "《高效液相色谱法测定马铃薯中α-茄碱含量》", 《食品科学》 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108827889A (en) * 2018-06-13 2018-11-16 江西中医药大学 A kind of glue class material discrimination method based on optical characteristics
CN109307638A (en) * 2018-08-07 2019-02-05 江苏大学 A kind of measuring method and device of steamed bun specific volume
CN110133049A (en) * 2019-04-10 2019-08-16 浙江大学 Tea grades fast non-destructive detection method based on electronic nose and machine vision
CN110702768A (en) * 2019-11-05 2020-01-17 广东省农业科学院农产品公共监测中心 Agricultural product quality production line dynamic lossless bionic olfaction detection equipment and method
CN111443160A (en) * 2020-01-23 2020-07-24 华东理工大学 Gas-sensitive-gas chromatography information fusion and electronic nose instrument on-line analysis method
CN111443160B (en) * 2020-01-23 2021-02-12 湖州老恒和酿造有限公司 Gas-sensitive-gas chromatography information fusion and electronic nose instrument on-line analysis method
WO2021147274A1 (en) * 2020-01-23 2021-07-29 华东理工大学 Gas-sensing-gas-phase electronic nose instrument and fermentation-malodor multi-state parameter online analysis method
CN112036470A (en) * 2020-08-28 2020-12-04 扬州大学 Cloud transmission-based multi-sensor fusion cucumber bemisia tabaci identification method
CN115578553A (en) * 2022-11-22 2023-01-06 河南知微生物工程有限公司 Method for quickly detecting formaldehyde based on time sequence image sequence

Similar Documents

Publication Publication Date Title
CN107748165A (en) Potato solanine detection method based on machine vision Yu electronic nose integration technology
Manavalan Automatic identification of diseases in grains crops through computational approaches: A review
Paymode et al. Transfer learning for multi-crop leaf disease image classification using convolutional neural network VGG
Khattak et al. Automatic detection of citrus fruit and leaves diseases using deep neural network model
Singh et al. A review of imaging techniques for plant disease detection
Yang et al. Diagnosis of plant cold damage based on hyperspectral imaging and convolutional neural network
CN101881726B (en) Nondestructive detection method for comprehensive character living bodies of plant seedlings
CN105181912B (en) A kind of Noninvasive Measuring Method of Freshness in rice storage
Mishra et al. A Deep Learning-Based Novel Approach for Weed Growth Estimation.
Valenzuela et al. Quality assessment of lettuce using artificial neural network
Yadav et al. A Novel Convolutional Neural Network Based Model for Recognition and Classification of Apple Leaf Diseases.
CN108020556A (en) A kind of online damage testing sorting technique of corn seed based on machine vision
Weng et al. Reflectance images of effective wavelengths from hyperspectral imaging for identification of Fusarium head blight-infected wheat kernels combined with a residual attention convolution neural network
Nagaraju et al. Transfer learning based convolutional neural network model for classification of mango leaves infected by anthracnose
Paymode et al. Tomato leaf disease detection and classification using convolution neural network
Thakur et al. Deep transfer learning based photonics sensor for assessment of seed-quality
Kalim et al. Citrus leaf disease detection using hybrid cnn-rf model
Mayuri et al. ROLE OF IMAGE PROCESSING AND MACHINE LEARNING TECHNIQUES IN DISEASE RECOGNITION, DIAGNOSIS AND YIELD PREDICTION OF CROPS: A REVIEW.
Sunitha et al. Modeling convolutional neural network for detection of plant leaf spot diseases
Alom et al. Species classification of brassica napus based on flowers, leaves, and packets using deep neural networks
Chauhan et al. Classification of Nutritional Deficiencies in Cabbage Leave Using Random Forest
Almalky et al. An Efficient Deep Learning Technique for Detecting and Classifying the Growth of Weeds on Fields
Lee et al. Development of image recognition and classification algorithm for tea leaf diseases using convolutional neural network
Benlachmi et al. Fruits Disease Classification using Machine Learning Techniques
CN115754184A (en) Rapid detection system and method for salmon earthy smell

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20180302