CN206788033U - A kind of dedicated test system of near infrared spectrum detection fruit internal quality - Google Patents
A kind of dedicated test system of near infrared spectrum detection fruit internal quality Download PDFInfo
- Publication number
- CN206788033U CN206788033U CN201720191816.6U CN201720191816U CN206788033U CN 206788033 U CN206788033 U CN 206788033U CN 201720191816 U CN201720191816 U CN 201720191816U CN 206788033 U CN206788033 U CN 206788033U
- Authority
- CN
- China
- Prior art keywords
- near infrared
- spectrometer
- sample
- detection
- infrared spectrum
- 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.)
- Expired - Fee Related
Links
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
A kind of dedicated test system of near infrared spectrum detection fruit internal quality, its composition are as follows:Spectrometer (10), fly to insult OK6410 A development boards (3), halogen light source (4), portable power source (5), fibre-optical probe (7), voltage reduction module (9), device outer container (1);Wherein:Power supply (5) is connected to halogen light source (4);It is 24V that power supply (5), which feeds voltage,;Halogen light source (4) is attached with detection probe (7) by circuit, thinks that detection probe (7) is powered;Power supply (5) is connected to spectrometer (10) by voltage reduction module (9);Spectrometer (10) is also connected with winged icepro OK 6410A development boards (3);Other devices are all disposed within device outer container (1).The utility model analyze speed is fast, and non-destructive is simple to operate, and cost is low, and it has expected more huge economy and social value.
Description
Technical field
The applied technical field that using near-infrared spectrum technique agricultural product are carried out with Non-Destructive Testing is the utility model is related to, it is special
A kind of dedicated test system of near infrared spectrum detection fruit internal quality is not provided.
Background technology
With the continuous improvement of living standard, it is outer not to be concerned only with size, color and profile etc. when buying fruit by consumer
Component matter, and higher requirement is it is also proposed to inside qualities such as mouthfeel, pol and acidity.At present, for fruit internal product
The detection of matter, mostly using artificial sense evaluation and the method for conventional chemical analysis.Artificial sense rating method, there is subjectivity
By force, the shortcomings of evaluation criterion differs and for conventional chemical analysis method, although it has, accuracy in detection is high and accuracy is strong
The advantages that, but its detection workload is big, detection time length, and sample need to be destroyed, these are not all suitable in fruit
The Fast nondestructive evaluation of component matter.
Domestic and foreign scholars have also carried out numerous studies for the design of portable apple detection means.Kang Zhi it is bright et al.
Utilize the irreflexive principle of visible ray near-infrared within 2010, design a kind of portable impaired fruit detection based on microprocessor
Device, realize identification and classification to being damaged apple and normal apple.The author is using red fuji apple as sample, by right
The analysis of test voltage, propose the detection method using standard deviation as threshold value.After tested, the device recognition correct rate reaches 90%
More than.Fan's book is auspicious et al. to be devised using ARM9 processors as core in 2014, is used as light using micro spectrometer and self-control hypocarp
Spectrum detection device, the portable apple sugar content spectrum detection instrument using WinCE as operating system.Research shows, the detector can be compared with
Meet the Fast nondestructive evaluation of apple sugar content well.The research provides for quick, portable apple sugar content spectrum detection instrument design
Reference.Guidetti et al. applied portable visible-near infrared spectrometer experimental system (wave-length coverage in 2010
450nm-980nm), modeled with reference to PLS PLS, the maturity of grape sample fresh fruit and homogeneous pulp is referred to
Mark (soluble solid (TSS), titrable acidity and pH value) and phenol mature indicator (anthocyanidin and polyphenol content) is carried out
Detection.Result of study shows:Portable visible-the near infrared spectrometer can be used as a kind of Fast nondestructive evaluation grape phase
Close the effective means of quality.U.S. Michael S.Wiederoder et al. developed portable EO-1 hyperion in 2011
Imaging system, for detecting the pollutant on food processing equipment, the device front end needs frame on tripod, battery, notebook
Computer and filter add the complexity of operation in rearmounted knapsack.
The highly desirable special inspection for obtaining a kind of excellent near infrared spectrum detection fruit internal quality of technique effect of people
Examining system.
Utility model content
The purpose of this utility model is to provide a kind of excellent near infrared spectrum detection fruit internal quality of technique effect
Dedicated test system.The defects of in order to overcome current domestic and international prior art, the utility model spy provide a kind of based near red
The cool humid zone quality damage-free detection method and system of external spectrum technology, key problem in technology are to apply Vis/NIR technology,
The visible and near infrared spectrum image information on tested cool humid zone surface can be quickly and accurately gathered, and it is analyzed, from
And realize to cool humid zone quality Fast nondestructive evaluation.
A kind of near infrared spectrum detection fruit internal quality method of the utility model, it is characterised in that:Initially set up and be based on
Near infrared spectrum detects the forecast model of fruit internal quality;Then near infrared spectrum detection method is used for fruit to be detected
Carry out detection and obtain detection data;Detection data are contrasted with the forecast model formerly prepared;Draw and be directed to water to be detected
The inside quality evaluation result of fruit;Wherein:The method for establishing the forecast model based near infrared spectrum detection fruit internal quality
Corresponding content requires as follows successively:
1) cool humid zone is gathered with spectrometer (10) to the various typical cool humid zone samples of known pol content first
The spectral information of sample surface, the spectral information of each sample collection at least three difference, is then averaging processing as each
Kind each self-corresponding spectrum representative information of typical sample, cool humid zone spectral information is cool humid zone in 485.01-900.71nm
Spectral reflectivity in wavelength under 1231 wavelength points;
2) cool humid zone sample spectrum data obtained by the step 1) are pre-processed;In original spectrum curve,
The both ends of wavelength, the noise that spectroscopic data is adulterated is more, and signal to noise ratio is relatively low, in order to improve the accuracy of established model, choosing
The spectral reflectance data under 485.01-900.71nm wavelength is taken to be pre-processed and modeled;It is visible near-infrared to collect sample
Spectroscopic data is imported in Unscrambler9.7 (chemo metric software of CAMO companies of the U.S.), using this software to spectrum number
According to progress de-noising pretreatment, specifically Moving Average (Moving average, MA) and polynomial convolution exponential smoothing respectively
(Savitzky-Golay, SG);Light path corrects pretreatment:Multiplicative scatter correction (Multiplicative Scatter
Correction, MSC), standard normal variable conversion (Standard Normal Variate Correction, SNV) and arrow
Amount normalization (Vector Normalization, VN)) and differential pre-process to obtain calibration set sample spectrum data;
3) information data in the pretreated calibration set sample spectrum of the step 2) is subjected to PLS point
Analysis;(PLS is multivariate statistics data analysing method, nineteen eighty-three A Banuo (C.Albano) and Wood (S.Wold) etc.
People proposes the algorithm for the first time;In recent years, it quickly grows in method and theory orientation, and application is more and more extensive;It can be incited somebody to action
The method of cognitive-ability and the method for modular form organically integrate, and under same algorithm, can complete data structure simultaneously
Correlation analysis (canonical correlation point between simplified (principal component analysis), regression modeling (multiple linear regression) and XY variables
Analysis), i.e. PLS ≈ multiple linear regressions+principal component analysis+canonical correlation analysis;PLS is that one kind is more
Dependent variable corresponds to the regression modeling method of more independents variable, the method that its application is all decomposed to variable X and Y respectively, exists first
Respective principal component (the commonly referred to as factor) is extracted respectively in variable Y and X, then by the factor of extraction according to correlation therebetween
Property size is arranged;So when establishing model, as long as being selected to factor number, it becomes possible to which participation models;Its formula
It is as follows:
X=TPT+E
Y=UQT+F
In formula, T=[t1, t2..., tn] and P=[p1, p2..., pn] be respectively X score matrix and loading matrix;U
=[u1, u2..., un] and Q=[q1, q2..., qn] be respectively Y score matrix and loading matrix;E=[e1, e2..., en
]TWith F=[f1, f2..., fn]TRespectively X and Y residual matrix;Associated between T and U by coefficient matrix B,
U=TB
B=(TTT)-1TTY
Matrix X for predicting sample*, score matrix T is obtained by loading matrix P*=X*P, pass through formula 2-12 and formula 2-13
Iteration can with draw prediction sample associated inner quality predicted value Y*For:
Y*=T*BQ=X*PBQ);
4) foundation of model:
According to the step 2), pretreated data application PLS 3) will be obtained it is modeled and records each
Model measurement index under model, most suitable spectroscopic data preprocess method are multiplicative scatter correction combination second-order differential;With
Modeling integrates 40 samples as research object, establishes the visible and near infrared spectrum forecast model of the pol of sample;Its pol model is public
Formula is:
Y=11.3628+ (- 9.2609) X1+0.0031X2+(-0.0284)X3+…+
(-0.0003)X1230+(-9.2609)X1231 (2)
The pol value of Y-apple sample, ° Brix in formula
X1…X1231Spectral reflectivity under 1231 wavelength points of-cool humid zone spectral information
Recycle this model to be predicted 20 samples of inspection set, obtain modeling collection, inspection set sample qualities measured value
With predicted value Comparative result;
The detected value of 20 apple samples and actual value contrast are as shown in Figure 6.
Absolute deviation distribution map is established according to the statistical result of the detected value of 20 apple samples and actual value, such as Fig. 7 institutes
Show.
From Fig. 6 and Fig. 7, the predicted value of most samples and the absolute deviation values of measured value within 0.15,
Absolute deviation values very little, meet the actual production requirement in fruits and vegetables detect, so this model can be applied to cool humid zone pol
It is actually detected;
1) forecast model is shown:
Model is established to the step 4) to analyze, gained forecast model precision of prediction is higher, and prediction is reliable, meets
Apple quality testing requirements, and can obtain cool humid zone pol combined prediction PLS (PLS i.e. combined predictions:PLS)
Forecast model expression formula, and finally complete the display of prediction result after input prediction collection sample spectral data.
The near infrared spectrum detects fruit internal quality method, it is characterised in that:For the various of known pol content
Typical cool humid zone sample, the spectral information that cool humid zone sample surface is gathered near infrared spectrometer obtain various samples
Also need to carry out following pretreatments to spectrum representative information after spectrum representative information:Cool humid zone spectroscopic data is imported into software
In Unscrambler9.7, the pretreatment of sample pol data is carried out, various preprocess methods are as shown in table 2:
The sample pol data preprocessing method statistical form of table 2
, it is necessary to have indices to the accuracy of model, reliability, stability and predictive ability after the foundation of mathematical modeling
Measurement and evaluation, coefficient of determination R are made etc. index2, root-mean-square error RMSEC, predicted root mean square error RMSEP are corrected, is deviateed
Rate Bias;Coefficient of determination R2It is higher, during RMSEP≤RMSEC, illustrate that institute's established model prediction effect is optimal;Product are learnt by the analysis of table 2
Prime number is micro- for multiplicative scatter correction combination second order according to optimal pretreatment method, the optimal spectrum data preprocessing method of apple sugar content
Divide (MSC+SD).
According to the near infrared spectrum detection fruit internal quality method of claim 1 or 2, it is characterised in that:It is described near
Infrared spectrum detection fruit internal quality method uses dedicated test system, and the composition of dedicated test system is as follows:
Mainly there is spectrometer 10, fly to insult OK6410-A development boards 3, halogen light source 4, portable power source 5, fibre-optical probe 7, drop
Die block 9, device outer container 1;Wherein:Power supply 5 is connected to halogen light source 4 to be powered for halogen light source 4;Power supply 5 is fed
Voltage is 24V;Halogen light source 4 is attached with detection probe 7 by circuit, thinks that detection probe 7 is powered;Power supply 5 is also logical
Cross voltage reduction module 9 and be connected to spectrometer 10 to be powered for spectrometer 10;Voltage reduction module 9 is specifically 5V 24V voltage conversions;
Spectrometer 10 is also connected with winged icepro OK-6410A development boards 3, to be powered to fly to insult OK6410-A development boards 3;Other devices are all
It is arranged on device outer container 1.
The dedicated test system of the near infrared spectrum detection fruit internal quality method also meets to require as follows:Spectrometer
10 be specially Vis/NIR instrument, is specifically purchased from U.S. ocean company, for being acquired to the spectral information of apple,
Using the second generation ILX511B CCD of Sony companies of Sony, (wherein, ILX511B CCD are the linear image sensing of 2048 pixels for it
Device), wave-length coverage 350-1000nm;Wavelength resolution is between 0.3-10nm, signal to noise ratio 250:1, most fast sample rate is
1ms;
USB2000+ spectrometers are that ARM11 processors are connected with flying to insult OK6410-A development boards 3 by USB data line interface;
Fly to insult the specifically ARM11 processors based on Samsung (SUMSUNG) company, while core of OK6410-A development boards 3
Plate ARM11 processors carry S3C6410 chips, and bottom plate is provided with various peripheral circuits, and S3C6410 is SUMSUNG (Samsung) public affairs
Take charge of the RISC (Reduced Instruction Set Computer) of the high-performance low-power-consumption of 16/32 based on ARM 1176
General processor, suitable for the terminal device such as movement, hand-held;S3C6410 peripheral circuit is mainly connect by usb circuit, SD
Mouth circuit, voltage conversion circuit, buzzer and LCD touch screen etc. are formed;Usb circuit is used to fly to insult OK6410-A exploitations
Communication between plate 3 and spectrometer 10, LCD touch screen are used for the control to whole detection process and the output of information, display, SD
Interface circuit is used for spectroscopic data and pol, the storage of acidity;
Halogen light source 4 is SPL-HL series tungsten halogen lamps, and its model SPL-HL-H, originating from Hangzhou spectrum radium photoelectric technology has
Limit company, it exports halogen tungsten lamp for a accurate fiber coupling, and its wave-length coverage covers visible ray-near infrared light (VIS-
NIR), wavelength 360-2500nm;The optical design of effectively optimizing causes tungstohalogen lamp to export advantage for low-power consumption high brightness, and
The mu balanced circuit design of optimization, it is ensured that the steady spectrum output of halogen tungsten lamp light source, its output stability is up to 0.2%;
The output voltage of portable power source 5 is 24V, and it is the power-on time about 2 hours of whole device, can be provided for light source surely
Fixed, permanent voltage, meets the needs of outdoor detection;
Detection probe used includes probe cover, optical fiber interface, optical fiber interface fixed seat three parts;
In addition, be pasted with fin on voltage reduction module 9, it is reached for that voltage reduction module 9 radiates and whole device system dissipates
The purpose of heat, and whole device outer container 1 is in open state, and so significantly whole device system can be radiated;
Tthe utility model system replaces the computer in typical detection system, and this practicality with winged icepro OK6410-A development boards
Novel system is powered with portable power source to whole device, and power-on time is longer, the use available for outdoor detection.
The dedicated test system of near infrared spectrum detection fruit internal quality is also claimed in the utility model, and its feature exists
In:The composition of the dedicated test system of the near infrared spectrum detection fruit internal quality is as follows:
Mainly have USB2000+ it is visible/spectrometer 10, fly to insult OK6410-A development boards 3, halogen light source 4, portable power source
5, fibre-optical probe 7, voltage reduction module 9, device outer container 1;Wherein:Power supply 5 is connected to halogen light source 4 to be halogen light source 4
Power supply;It is 24V that power supply 5, which feeds voltage,;Halogen light source 4 is attached with detection probe 7 by circuit, thinks detection probe 7
Power supply;Power supply 5 is also connected to spectrometer 10 to be powered for spectrometer 10 by voltage reduction module 9;The specifically handle of voltage reduction module 9
24V voltage conversions are 5V;Spectrometer 10 is also connected with winged icepro OK-6410A development boards 3, to insult OK6410-A development boards 3 to be winged
Power supply;Other devices are all disposed within device outer container 1.
The dedicated test system of the near infrared spectrum detection fruit internal quality also meets to require as follows:Spectrometer 10 has
Body is Vis/NIR instrument, is specifically purchased from U.S. ocean company, for being acquired to the spectral information of apple, it is adopted
With the second generation ILX511B CCD (wherein, ILX511B CCD are 2048 pixel linear imaging sensors) of Sony companies of Sony,
Wave-length coverage 350-1000nm;Wavelength resolution is between 0.3-10nm, signal to noise ratio 250:1, most fast sample rate is 1ms;
USB2000+ spectrometers are that ARM11 processors are connected with flying to insult OK6410-A development boards 3 by USB data line interface;
Fly to insult the specifically ARM11 processors based on Samsung (SUMSUNG) company, while core of OK6410-A development boards 3
Plate ARM11 processors carry S3C6410 chips, and bottom plate is provided with various peripheral circuits, and S3C6410 is SUMSUNG (Samsung) public affairs
Take charge of the RISC (Reduced Instruction Set Computer) of the high-performance low-power-consumption of 16/32 based on ARM 1176
General processor, suitable for the terminal device such as movement, hand-held;S3C6410 peripheral circuit is mainly connect by usb circuit, SD
Mouth circuit, voltage conversion circuit, buzzer and LCD touch screen etc. are formed;Usb circuit is used to fly to insult OK6410-A exploitations
Communication between plate 3 and spectrometer 10, LCD touch screen are used for the control to whole detection process and the output of information, display, SD
Interface circuit is used for spectroscopic data and pol, the storage of acidity;
Halogen light source 4 is SPL-HL series tungsten halogen lamps, and its model SPL-HL-H, originating from Hangzhou spectrum radium photoelectric technology has
Limit company, it exports halogen tungsten lamp for a accurate fiber coupling, and its wave-length coverage covers visible ray-near infrared light (VIS-
NIR), wavelength 360-2500nm;The optical design of effectively optimizing causes tungstohalogen lamp to export advantage for low-power consumption high brightness, and
The mu balanced circuit design of optimization, it is ensured that the steady spectrum output of halogen tungsten lamp light source, its output stability is up to 0.2%;
The output voltage of portable power source 5 is 24V, and it is the power-on time about 2 hours of whole device, can be provided for light source surely
Fixed, permanent voltage, meets the needs of outdoor detection;
Detection probe used includes probe cover, optical fiber interface, optical fiber interface fixed seat three parts;
In addition, be pasted with fin on voltage reduction module 9, it is reached for that voltage reduction module 9 radiates and whole device system dissipates
The purpose of heat, and whole device outer container 1 is in open state, and so significantly whole device system can be radiated;
Tthe utility model system replaces the computer in typical detection system, and this practicality with winged icepro OK6410-A development boards
Novel system is powered with portable power source to whole device, and power-on time is longer, the use available for outdoor detection.
As shown in Figure 1, Figure 3, the operation principle of Tthe utility model system is:Cool humid zone sample is placed on fibre-optical probe 7
Test position, the detected sample of the irradiation of halogen light source 4, the light passed through from sample is transmitted to through fibre-optical probe 7 and optical fiber
Spectrometer 10, spectrometer convey data to the embedded main board processing systems of ARM 11 by serial ports and handled, and by result
It is shown on LED touch screen.
As shown in Fig. 2 its detecting system course of work is:It is first clear to the cool humid zone sample just bought before starting detection
Wash and write apple sample message and the place of production, then device for opening power supply, activation system hardware (development board, light source), light source is pre-
Heat 3 minutes.The spectral intensity of light source starts to detect after reaching stable, meanwhile, select arrange parameter (smoothness, the product of spectrometer
Between timesharing, sampling number).Then detection probe close to cool humid zone sample surfaces, then push interface starts to detect
Button, spectrometer collection cool humid zone sample spectral data, and the data input detected is carried out to embedded mathematical modeling
Forecast analysis, detection parameters and class information are returned, is output on interface.
The utility model proposes a kind of near infrared spectrum Non-Destructive Testing cool humid zone inside quality software systems, be
In WinCE6.0 systems embedded ARM11, using VS2005 programming tools, the software systems of hommization are designed, complete spectrum
Instrument parameter sets, gathers half-light spectrum, removes half-light spectrum, storage reference spectra, sample spectral data gathers and storage, and collection
To spectroscopic data be loaded into software model equation, calculate the pol value of sample, and testing result is shown on screen.
Software write, debug it is errorless after, into mancarried device ARM11 plates, it is soft under WinCE6.0 systems to run this for cross compile
Part, its effect are as shown in Figure 5.
The utility model is from " cool humid zone " for being purchased from Yuan Nei fruit supermarket of Agricultural University Of Shenyang, to ensure the generation of sample
Table and rich, it is unified to choose that no significant defect, color and luster be unified, storage time is closer to, selected directly using slide measure
Apple sample of the footpath scope between 70-100mm.
The pol in refractive power saccharimeter measurement sample pol determination sample is held using the portable RHB-080 types of Taiwan LINK
Content, the calibration set using this batch of sample as founding mathematical models carry out founding mathematical models, it is proposed that one kind utilizes cool humid zone
Near infrared spectrum in the method comprising the main component and the information measurement of measurement of sample its pol content, this method application
Learn metrology method and research is associated to cool humid zone near infrared spectrum and pol content, it may be determined that determine between both
Property or quantitative relationship, i.e. forecast model.
After establishing forecast model, as long as measuring the near infrared spectrum of unknown sample, it is assured that according to forecast model
The pol content of cool humid zone.This method has that analyze speed is fast, analysis efficiency is high, without using any chemical reagents, is parsed into
This is low, and the advantages of do not cause any pollution to environment.
It is one of focus of domestic and international agricultural product scientific research using Vis/NIR technology for detection fruit quality,
One typical detecting system is generally made up of camera bellows, spectrometer, computer, light source, detection probe etc..It is related in the prior art
Detecting system generally can have that volume is big, is not easy in field operation using computer as processing platform, mobile inconvenience etc.
Shortcoming.
And the utility model is based on embedded system and Vis/NIR detection technique, with ARM11 (advanced
RISC machines) processor is core, using halogen light source 4 and USB2000+ Vis/NIRs instrument as spectral detection
The core of device, in the WinCE6.0 systems that ARM11 is embedded, using vs2005 programming tools, design man-machine interaction
Interface;Meanwhile replace laboratory computer to carry out data processing using winged icepro OK6410-A development boards, utilize portable power source pair
Its whole device system carries out long-time power supply, the detecting system volume is significantly reduced, can be reached in outdoor long-time operation
To embedded and portable purpose.
The utility model is based on near-infrared spectrum technique, have analyze speed is fast, non-destructive, sample without pretreatment,
It can realize to simultaneous determination of multiponents, and low cost and simple operation and other advantages.It has expected more huge
Economic value and social value.
Brief description of the drawings
Below in conjunction with the accompanying drawings and embodiment is described in further detail to the utility model:
Fig. 1 is detecting system operation principle block diagram;
Fig. 2 is cool humid zone overhaul flow chart;
Fig. 3 is detecting system hardware principle of compositionality simplified schematic diagram;
Fig. 4 device software system interfaces;
Fig. 5 is sample Vis/NIR Spectroscopy diffusing reflection rate spectrogram;
Fig. 6 is the pol detected value and actual value comparison diagram of 20 apple samples;
Fig. 7 is the absolute deviation distribution map according to the statistical result foundation of the detected value and actual value of 20 apple samples.
Embodiment
Reference implication is as follows:1- outer containers;2-LED touch-screens;3- development board mainboards;4- halogen light sources;5- power supplys;
6- fin;7- detection probes;8- optical fiber;9- voltage reduction modules;10- near infrared spectrometers.
Embodiment 1
A kind of near infrared spectrum detects fruit internal quality method:Initially set up based near infrared spectrum detection fruit internal
The forecast model of quality;Then carry out detection using near infrared spectrum detection method for fruit to be detected and obtain detection data;
Detection data are contrasted with the forecast model formerly prepared;Draw the inside quality evaluation result for fruit to be detected;
Wherein:Content corresponding to establishing the method for the forecast model based near infrared spectrum detection fruit internal quality requires such as successively
Under:
1) cool humid zone is gathered with spectrometer (10) to the various typical cool humid zone samples of known pol content first
The spectral information of sample surface, the spectral information of each sample collection at least three difference, is then averaging processing as each
Kind each self-corresponding spectrum representative information of typical sample, cool humid zone spectral information is cool humid zone in 485.01-900.71nm
Spectral reflectivity in wavelength under 1231 wavelength points;
2) cool humid zone sample spectrum data obtained by the step 1) are pre-processed;In original spectrum curve,
The both ends of wavelength, the noise that spectroscopic data is adulterated is more, and signal to noise ratio is relatively low, in order to improve the accuracy of established model, choosing
The spectral reflectance data under 485.01-900.71nm wavelength is taken to be pre-processed and modeled;It is visible near-infrared to collect sample
Spectroscopic data is imported in Unscrambler9.7 (chemo metric software of CAMO companies of the U.S.), using this software to spectrum number
According to progress de-noising pretreatment, specifically Moving Average (Moving average, MA) and polynomial convolution exponential smoothing respectively
(Savitzky-Golay, SG);Light path corrects pretreatment:Multiplicative scatter correction (Multiplicative Scatter
Correction, MSC), standard normal variable conversion (Standard Normal Variate Correction, SNV) and arrow
Amount normalization (Vector Normalization, VN)) and differential pre-process to obtain calibration set sample spectrum data;
3) information data in the pretreated calibration set sample spectrum of the step 2) is subjected to PLS point
Analysis;(PLS is multivariate statistics data analysing method, nineteen eighty-three A Banuo (C.Albano) and Wood (S.Wold) etc.
People proposes the algorithm for the first time;In recent years, it quickly grows in method and theory orientation, and application is more and more extensive;It can be incited somebody to action
The method of cognitive-ability and the method for modular form organically integrate, and under same algorithm, can complete data structure simultaneously
Correlation analysis (canonical correlation point between simplified (principal component analysis), regression modeling (multiple linear regression) and XY variables
Analysis), i.e. PLS ≈ multiple linear regressions+principal component analysis+canonical correlation analysis;PLS is that one kind is more
Dependent variable corresponds to the regression modeling method of more independents variable, the method that its application is all decomposed to variable X and Y respectively, exists first
Respective principal component (the commonly referred to as factor) is extracted respectively in variable Y and X, then by the factor of extraction according to correlation therebetween
Property size is arranged;So when establishing model, as long as being selected to factor number, it becomes possible to which participation models;Its formula
It is as follows:
X=TPT+E
Y=UQT+F
In formula, T=[t1, t2..., tn] and P=[p1, p2..., pn] be respectively X score matrix and loading matrix;U
=[u1, u2..., un] and Q=[q1, q2..., qn] be respectively Y score matrix and loading matrix;E=[e1, e2..., en
]TWith F=[f1, f2..., fn]TRespectively X and Y residual matrix;Associated between T and U by coefficient matrix B,
U=TB
B=(TTT)-1TTY
Matrix X for predicting sample*, score matrix T is obtained by loading matrix P*=X*P, pass through formula 2-12 and formula 2-13
Iteration can with draw prediction sample associated inner quality predicted value Y*For:
Y*=T*BQ=X*PBQ);
4) foundation of model:
According to the step 2), pretreated data application PLS 3) will be obtained it is modeled and records each
Model measurement index under model, most suitable spectroscopic data preprocess method are multiplicative scatter correction combination second-order differential;With
Modeling integrates 40 samples as research object, establishes the visible and near infrared spectrum forecast model of the pol of sample;Its pol model is public
Formula is:
Y=11.3628+ (- 9.2609) X1+0.0031X2+(-0.0284)X3+…+
(-0.0003)X1230+(-9.2609)X1231 (2)
The pol value of Y-apple sample, ° Brix in formula
X1…X1231Spectral reflectivity under 1231 wavelength points of-cool humid zone spectral information
Recycle this model to be predicted 20 samples of inspection set, obtain modeling collection, inspection set sample qualities measured value
With predicted value Comparative result;
The detected value of 20 apple samples and actual value contrast are as shown in Figure 6.
Absolute deviation distribution map is established according to the statistical result of the detected value of 20 apple samples and actual value, such as Fig. 7 institutes
Show.
From Fig. 6 and Fig. 7, the predicted value of most samples and the absolute deviation values of measured value within 0.15,
Absolute deviation values very little, meet the actual production requirement in fruits and vegetables detect, so this model can be applied to cool humid zone pol
It is actually detected;
5) forecast model is shown:
Model is established to the step 4) to analyze, gained forecast model precision of prediction is higher, and prediction is reliable, meets
Apple quality testing requirements, and can obtain cool humid zone pol combined prediction PLS (PLS i.e. combined predictions:PLS)
Forecast model expression formula, and finally complete the display of prediction result after input prediction collection sample spectral data.
For the various typical cool humid zone samples of known pol content, cool humid zone sample is gathered near infrared spectrometer
The spectral information on this surface also needs to carry out following pre- places to spectrum representative information after obtaining the spectrum representative information of various samples
Reason:Cool humid zone spectroscopic data is imported in software Unscrambler9.7, carries out the pretreatment of sample pol data, it is various pre-
Processing method is as shown in table 2:
The sample pol data preprocessing method statistical form of table 2
, it is necessary to have indices to the accuracy of model, reliability, stability and predictive ability after the foundation of mathematical modeling
Measurement and evaluation, coefficient of determination R are made etc. index2, root-mean-square error RMSEC, predicted root mean square error RMSEP are corrected, is deviateed
Rate Bias;Coefficient of determination R2It is higher, during RMSEP≤RMSEC, illustrate that institute's established model prediction effect is optimal;Product are learnt by the analysis of table 2
Prime number is micro- for multiplicative scatter correction combination second order according to optimal pretreatment method, the optimal spectrum data preprocessing method of apple sugar content
Divide (MSC+SD).
The near infrared spectrum detection fruit internal quality method uses dedicated test system, the composition of dedicated test system
It is as follows:
Mainly there is spectrometer 10, fly to insult OK6410-A development boards 3, halogen light source 4, portable power source 5, fibre-optical probe 7, drop
Die block 9, device outer container 1;Wherein:Power supply 5 is connected to halogen light source 4 to be powered for halogen light source 4;Power supply 5 is fed
Voltage is 24V;Halogen light source 4 is attached with detection probe 7 by circuit, thinks that detection probe 7 is powered;Power supply 5 is also logical
Cross voltage reduction module 9 and be connected to spectrometer 10 to be powered for spectrometer 10;Voltage reduction module 9 is specifically 5V 24V voltage conversions;
Spectrometer 10 is also connected with winged icepro OK-6410A development boards 3, to be powered to fly to insult OK6410-A development boards 3;Other devices are all
It is arranged on device outer container 1.
The dedicated test system of the near infrared spectrum detection fruit internal quality method also meets to require as follows:Spectrometer
10 be specially Vis/NIR instrument, is specifically purchased from U.S. ocean company, for being acquired to the spectral information of apple,
Using the second generation ILX511B CCD of Sony companies of Sony, (wherein, ILX511B CCD are the linear image sensing of 2048 pixels for it
Device), wave-length coverage 350-1000nm;Wavelength resolution is between 0.3-10nm, signal to noise ratio 250:1, most fast sample rate is
1ms;
USB2000+ spectrometers are that ARM11 processors are connected with flying to insult OK6410-A development boards 3 by USB data line interface;
Fly to insult the specifically ARM11 processors based on Samsung (SUMSUNG) company, while core of OK6410-A development boards 3
Plate ARM11 processors carry S3C6410 chips, and bottom plate is provided with various peripheral circuits, and S3C6410 is SUMSUNG (Samsung) public affairs
Take charge of the RISC (Reduced Instruction Set Computer) of the high-performance low-power-consumption of 16/32 based on ARM 1176
General processor, suitable for the terminal device such as movement, hand-held;S3C6410 peripheral circuit is mainly connect by usb circuit, SD
Mouth circuit, voltage conversion circuit, buzzer and LCD touch screen etc. are formed;Usb circuit is used to fly to insult OK6410-A exploitations
Communication between plate 3 and spectrometer 10, LCD touch screen are used for the control to whole detection process and the output of information, display, SD
Interface circuit is used for spectroscopic data and pol, the storage of acidity;
Halogen light source 4 is SPL-HL series tungsten halogen lamps, and its model SPL-HL-H, originating from Hangzhou spectrum radium photoelectric technology has
Limit company, it exports halogen tungsten lamp for a accurate fiber coupling, and its wave-length coverage covers visible ray-near infrared light (VIS-
NIR), wavelength 360-2500nm;The optical design of effectively optimizing causes tungstohalogen lamp to export advantage for low-power consumption high brightness, and
The mu balanced circuit design of optimization, it is ensured that the steady spectrum output of halogen tungsten lamp light source, its output stability is up to 0.2%;
The output voltage of portable power source 5 is 24V, and it is the power-on time about 2 hours of whole device, can be provided for light source surely
Fixed, permanent voltage, meets the needs of outdoor detection;
Detection probe used includes probe cover, optical fiber interface, optical fiber interface fixed seat three parts;
In addition, be pasted with fin on voltage reduction module 9, it is reached for that voltage reduction module 9 radiates and whole device system dissipates
The purpose of heat, and whole device outer container 1 is in open state, and so significantly whole device system can be radiated;
The present embodiment system replaces the computer in typical detection system, and the present embodiment with winged icepro OK6410-A development boards
System is powered with portable power source to whole device, and power-on time is longer, the use available for outdoor detection.
As shown in Figure 1, Figure 3, the operation principle of the present embodiment system is:Cool humid zone sample is placed on fibre-optical probe 7
Test position, the detected sample of the irradiation of halogen light source 4, the light passed through from sample are transmitted to light through fibre-optical probe 7 and optical fiber
Spectrometer 10, spectrometer conveys data to the embedded main board processing systems of ARM 11 by serial ports and handled, and result is shown
Show on LED touch screen.
As shown in Fig. 2 its detecting system course of work is:It is first clear to the cool humid zone sample just bought before starting detection
Wash and write apple sample message and the place of production, then device for opening power supply, activation system hardware (development board, light source), light source is pre-
Heat 3 minutes.The spectral intensity of light source starts to detect after reaching stable, meanwhile, select arrange parameter (smoothness, the product of spectrometer
Between timesharing, sampling number).Then detection probe close to cool humid zone sample surfaces, then push interface starts to detect
Button, spectrometer collection cool humid zone sample spectral data, and the data input detected is carried out to embedded mathematical modeling
Forecast analysis, detection parameters and class information are returned, is output on interface.
The present embodiment propose a kind of near infrared spectrum Non-Destructive Testing cool humid zone inside quality software systems, be
In WinCE6.0 systems embedded ARM11, using VS2005 programming tools, the software systems of hommization are designed, complete spectrum
Instrument parameter sets, gathers half-light spectrum, removes half-light spectrum, storage reference spectra, sample spectral data gathers and storage, and collection
To spectroscopic data be loaded into software model equation, calculate the pol value of sample, and testing result is shown on screen.
Software write, debug it is errorless after, into mancarried device ARM11 plates, it is soft under WinCE6.0 systems to run this for cross compile
Part, its effect are as shown in Figure 5.
The present embodiment is from " cool humid zone " for being purchased from Yuan Nei fruit supermarket of Agricultural University Of Shenyang, to ensure the representative of sample
Property and it is rich, it is unified to choose that no significant defect, color and luster be unified, storage time is closer to, select diameter using slide measure
Apple sample of the scope between 70-100mm.
The pol in refractive power saccharimeter measurement sample pol determination sample is held using the portable RHB-080 types of Taiwan LINK
Content, the calibration set using this batch of sample as founding mathematical models carry out founding mathematical models, it is proposed that one kind utilizes cool humid zone
Near infrared spectrum in the method comprising the main component and the information measurement of measurement of sample its pol content, this method application
Learn metrology method and research is associated to cool humid zone near infrared spectrum and pol content, it may be determined that determine between both
Property or quantitative relationship, i.e. forecast model.
After establishing forecast model, as long as measuring the near infrared spectrum of unknown sample, it is assured that according to forecast model
The pol content of cool humid zone.This method has that analyze speed is fast, analysis efficiency is high, without using any chemical reagents, is parsed into
This is low, and the advantages of do not cause any pollution to environment.
It is one of focus of domestic and international agricultural product scientific research using Vis/NIR technology for detection fruit quality,
One typical detecting system is generally made up of camera bellows, spectrometer, computer, light source, detection probe etc..It is related in the prior art
Detecting system generally can have that volume is big, is not easy in field operation using computer as processing platform, mobile inconvenience etc.
Shortcoming.
And the present embodiment is based on embedded system and Vis/NIR detection technique, with ARM11 (advanced
RISC machines) processor is core, using halogen light source 4 and USB2000+ Vis/NIRs instrument as spectral detection
The core of device, in the WinCE6.0 systems that ARM11 is embedded, using vs2005 programming tools, design man-machine interaction
Interface;Meanwhile replace laboratory computer to carry out data processing using winged icepro OK6410-A development boards, utilize portable power source pair
Its whole device system carries out long-time power supply, the detecting system volume is significantly reduced, can be reached in outdoor long-time operation
To embedded and portable purpose.
The present embodiment is based on near-infrared spectrum technique, have analyze speed is fast, non-destructive, sample without pretreatment, can
To realize to simultaneous determination of multiponents, and low cost and simple operation and other advantages.It has expected more huge warp
Ji value and social value.
Embodiment 2
1. test material:Test specimen is " cool humid zone ", originates from Shenyang City, Liaoning Province, picks up from Agricultural University Of Shenyang's scientific research
Experimental Base.Sample size is 60, unified to choose no significant defect, color and luster system to ensure the representative and rich of sample
One apple sample.Sample of the diameter range between 70-100mm is selected using slide measure, 60 samples are divided into modeling
Collection 40 and inspection set 20.Harvesting the same day take back laboratory, clean and is dried, number, mark after use 0.2mm polyethylene modeling
Expect preservative film packaging, be placed in (0 ± 0.5) DEG C refrigerating chamber store it is standby.Taken out before experiment from refrigerating chamber, place 5h, make sample
This bulk temperature is consistent with experimental enviroment (temperature is 21 DEG C, relative humidity 19%).
2. spectrum data gathering:Detecting system power supply is opened, is preheated 3 minutes, system detectio parameter is set, and the time of integration is
100ms, average time 2, smoothness 6.The black reference plate of standard correction is placed in detection probe, collection half-light spectrum, so
Standard correction blank is placed afterwards, gathers reference spectra.Detection probe is placed on apple surface again, cool humid zone sample is carried out
Collection, in order to obtain more representative information, reply apple male and female face is all irradiated and is acquired in proportion.Each sample
Then the representative information as the spectrum, sample visible/near infrared is averaging processing in the spectral information of 3 differences of this collection
Light diffusing reflection rate spectrogram is as shown in Figure 5.Spectral reflectivity calculation formula is:
In formula:RλThe apple sample spectra reflectivity under-wavelength X;
SλThe spectral intensity that the apple sample is reflected under-wavelength X;
DλThe spectral intensity of-half-light spectrum;
RWλThe spectral intensity of-reference spectra.
3. apple quality physico-chemical process is demarcated:Apple quality includes fruit shape size, color, smell, nutriment and contained
Several indexs such as amount, browning degree, edible rate.Pol is mainly chosen in experiment as cool humid zone chemical quality evaluation index.
According to standard GB/T/T10651-2008 (fresh apple), refractive power saccharimeter is held using the portable RHB-080 types of Taiwan LINK
Measure sample pol.Sampling point is sample spectra acquisition region in experiment.Measurement result is as shown in table 1.
The statistical form of the cool humid zone saccharic acid brittleness chemical score of table 1
4. spectroscopic data pre-processes
In original spectrum curve, at the both ends of wavelength, the noise that spectroscopic data is adulterated is more, and signal to noise ratio is relatively low, is
The accuracy of model will be established by improving, and the spectral reflectance data chosen under 485.01-900.71nm wavelength is pre-processed
And modeling.Collect sample visible and near infrared spectrum data to import in Unscrambler9.7 (CAMO companies of the U.S.), quantitative point
Spectral information is analysed, it is determined that being adapted to the preprocess method of apple quality information, establishes pol mathematical modeling.
For sample pol, different spectroscopic data preprocess methods, obtained PLS (PLS) mathematical modeling
As a result it is different.Comprehensive each model-evaluation index result, find the spectroscopic data pretreatment of multiplicative scatter correction combination second-order differential
The method coefficient of determination (0.9908) is higher, and correction root-mean-square error (0.1242) and predicted root mean square error (0.1189) are relatively
It is small, and RMSEP (0.1189)<RMSEC(0.1242).Illustrate for pol quality, most suitable spectroscopic data preprocess method
For multiplicative scatter correction combination second-order differential.
The sample pol data preprocessing method statistical form of table 2
Sample pol quality prediction model, coefficient of determination R2=0.9852, predicted root mean square error RMSEP (0.0958)<
Correct root-mean-square error RMSEC (0.1172), deviation Bias=-0.0133.Understanding, the forecast model prediction result is reasonable,
Precision of prediction relative ideal.Experiment integrates 40 samples as research object to model, and establishes the pol Vis/NIR of sample
Forecast model, this model is recycled to 20 sample predictions of inspection set.
The inspection set sample pol actual value of table 3 and predicted value statistical form
The predicted value of most samples and the absolute deviation values of measured value are within 0.15, absolute deviation values very little,
Meet the actual production requirement in fruits and vegetables detection, so this model can be applied to the actually detected of cool humid zone pol.
Claims (2)
- A kind of 1. dedicated test system of near infrared spectrum detection fruit internal quality, it is characterised in that:The near infrared spectrum The composition for detecting the dedicated test system of fruit internal quality is as follows:Spectrometer (10), fly to insult OK6410-A development boards (3), halogen light source (4), portable power source (5), fibre-optical probe (7), drop Die block (9), device outer container (1);Wherein:Power supply (5) is connected to halogen light source (4);It is 24V that power supply (5), which feeds voltage,;Halogen Plain lamp source (4) is attached with detection probe (7) by circuit, thinks that detection probe (7) is powered;Power supply (5) also passes through drop Die block (9) is connected to spectrometer (10);Spectrometer (10) is also connected with winged icepro OK-6410A development boards (3);Other devices are all It is arranged on device outer container (1).
- 2. the dedicated test system of fruit internal quality is detected according to the near infrared spectrum described in claim 1, it is characterised in that: The dedicated test system of the near infrared spectrum detection fruit internal quality also meets to require as follows:Spectrometer (10) is specially can / near infrared spectrometer is seen, using the pixel linear imaging sensor ILX511B CCD of the second generation 2048 of Sony, wavelength model Enclose 350-1000nm;Wavelength resolution is between 0.3-10nm, signal to noise ratio 250:1, most fast sample rate is 1ms;USB2000+ spectrometers are that ARM11 processors are connected with flying to insult OK6410-A development boards (3) by USB data line interface;Fly to insult OK6410-A development boards (3) the specifically ARM11 processors based on Samsung, while core board ARM11 processing Device carries S3C6410 chips, and S3C6410 peripheral circuit is by usb circuit, SD interface circuit, voltage conversion circuit, buzzing Device and LCD touch screen are formed;Halogen light source (4) is SPL-HL series tungsten halogen lamps, its model SPL-HL-H, it is limited to originate from Hangzhou spectrum radium photoelectric technology Company;Portable power source (5) is the electric supply installation that output voltage is 24V;Detection probe used includes probe cover, optical fiber interface, optical fiber interface fixed seat three parts;In addition, it is pasted with fin on voltage reduction module (9).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201720191816.6U CN206788033U (en) | 2017-03-01 | 2017-03-01 | A kind of dedicated test system of near infrared spectrum detection fruit internal quality |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201720191816.6U CN206788033U (en) | 2017-03-01 | 2017-03-01 | A kind of dedicated test system of near infrared spectrum detection fruit internal quality |
Publications (1)
Publication Number | Publication Date |
---|---|
CN206788033U true CN206788033U (en) | 2017-12-22 |
Family
ID=60703825
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201720191816.6U Expired - Fee Related CN206788033U (en) | 2017-03-01 | 2017-03-01 | A kind of dedicated test system of near infrared spectrum detection fruit internal quality |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN206788033U (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106841103A (en) * | 2017-03-01 | 2017-06-13 | 沈阳农业大学 | Near infrared spectrum detects fruit internal quality method and dedicated test system |
CN108303379A (en) * | 2018-01-08 | 2018-07-20 | 中国农业大学 | The potato detection device of hand-held |
-
2017
- 2017-03-01 CN CN201720191816.6U patent/CN206788033U/en not_active Expired - Fee Related
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106841103A (en) * | 2017-03-01 | 2017-06-13 | 沈阳农业大学 | Near infrared spectrum detects fruit internal quality method and dedicated test system |
CN108303379A (en) * | 2018-01-08 | 2018-07-20 | 中国农业大学 | The potato detection device of hand-held |
CN108303379B (en) * | 2018-01-08 | 2024-03-15 | 中国农业大学 | Handheld potato detection equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106841103A (en) | Near infrared spectrum detects fruit internal quality method and dedicated test system | |
CN101308086B (en) | Fruit internal quality on-line checking apparatus based on near infrared spectra technology | |
CN101907564B (en) | Rapeseed quality non-destructive testing method and device based on near infrared spectrum technology | |
CN105486655B (en) | The soil organism rapid detection method of model is intelligently identified based on infrared spectroscopy | |
CN104897607A (en) | Food modeling and rapid detecting integration method and system adopting portable NIRS (near infrared spectroscopy) | |
Xudong et al. | Nondestructive assessment of quality of Nanfeng mandarin fruit by a portable near infrared spectroscopy | |
CN105092579A (en) | Mango quality non-destructive testing device | |
CN105527246A (en) | Portable near infrared spectrum detection system | |
CN111537469A (en) | Apple quality rapid nondestructive testing method based on near-infrared technology | |
CN206788033U (en) | A kind of dedicated test system of near infrared spectrum detection fruit internal quality | |
CN101614663A (en) | As seen and near infrared spectrum bee pollen variety discriminating | |
CN101349638A (en) | Spectrum rapid nondestructive testing method for vitamin C content of fruits and vegetables | |
CN205719951U (en) | Detect black box and utilize the lossless quick detection equipment of Apple Mould Core of this detection black box | |
CN105527236A (en) | Method for determination of main nutritional components of agricultural product by use of spectroscopy method | |
CN107132197B (en) | Method and device for detecting total acid content of vinegar | |
CN105784606A (en) | Optical property based water quality monitoring system | |
CN113030011A (en) | Rapid nondestructive testing method and system for sugar content of fruits | |
CN207779882U (en) | A kind of peanut quality non-destructive testing device | |
CN201724900U (en) | Rapeseed quality nondestructive testing device based on near infrared spectrum technique | |
CN101858862B (en) | Method and system for quickly detecting thioglycoside content of rape seeds | |
CN105158178A (en) | Rapid modeling method for detecting sugar content of navel orange based on spectral peak area in high spectral transmission technology | |
CN201072405Y (en) | Spectrum rapid nondestructive testing device for vitamin C content of fruits and vegetables | |
CN104181125A (en) | Method for rapidly determining Kol-bach value of beer malt | |
CN209802978U (en) | remote real-time online fruit quality monitoring system | |
Luo et al. | Nondestructive measurement of sugar content in navel orange based on Vis-NIR spectroscopy |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20171222 Termination date: 20190301 |
|
CF01 | Termination of patent right due to non-payment of annual fee |