CN105427274B - It is a kind of that rotten citrus image detecting method is caused by mould infection - Google Patents
It is a kind of that rotten citrus image detecting method is caused by mould infection Download PDFInfo
- Publication number
- CN105427274B CN105427274B CN201510714053.4A CN201510714053A CN105427274B CN 105427274 B CN105427274 B CN 105427274B CN 201510714053 A CN201510714053 A CN 201510714053A CN 105427274 B CN105427274 B CN 105427274B
- Authority
- CN
- China
- Prior art keywords
- image
- citrus
- thumbnail
- wavelength
- characteristic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The present invention proposes that a kind of infected by mould causes rotten citrus image detecting method, including step：A gathers the Single wavelength spectrum picture of 4 characteristic wave strong points of citrus fruit to be measured, and B carries out the uneven correction of surface brightness to image；C：Superposition obtains characteristic wavelength combination image I after correction；D characteristic wavelength combination images I is converted to thumbnail F1；The R component image of E extraction thumbnail F1 three primary colours components simultaneously carries out image preprocessing to it；F is again converted to thumbnail F2 and extracts the R component image of thumbnail three primary colours component；G carries out target area segmentation to the R component image obtained in step F, carries out circularity judgement.Method proposed by the present invention, by the Single wavelength spectrum picture for gathering 4 characteristic wave strong points of citrus fruit to be measured, citrus fungal infection region inconspicuous is realized with reference to image combinatorial formula and image processing method and clearly visualizes display, so as to which the citrus that rotted caused by mould infection be identified more effectively, exactly.
Description
Technical field
The invention belongs to technical field of spectral detection, and in particular to a kind of detection method based on visible-near-infrared spectrum.
Background technology
Citrus is unique flavor, nutritious fruit, is well received by consumers.China is the big citrus life of the first in the world
State is produced, Citrus Industry all has huge economic market in the international or country, existed to lift citrus fruit quality
Domestic and international market competitiveness, automation rapid classification technology receives much concern always after the adopting of citrus fruit.With it is other often
See that External Defect such as scar is compared, the defects of being citrus fruit most serious of being rotted as caused by fungal infection, fungal disease infects
It is to cause the main reason that fresh citrus is rotten in storage and transport process, Penicillium digitatum Penicillium digitatum infection is outstanding
To be important.Research in recent years finds that China's citrus postharvest decay surpasses 80% as caused by infecting Penicillium digitatum.
At present, the computer vision technique based on RGB color camera has been used to the detection of citrus External Defect.However,
The incipient decay fruit caused by Penicillium digitatum infects, its infected zone coat color and normal fruit colour are about the same, lead
Cause extremely difficult using traditional this defect of RGB color camera calibration.Blasco et al. is detected using traditional RGB vision systems
11 type citrus skin defects including incipient decay caused by Penicillium digitatum infection, research find the system to other
The discrimination of type skin defects is very high, and for incipient decay fruit caused by Penicillium digitatum, its accuracy of detection only obtains
56.5% (Blasco, J., Aleixos, N., G ó mez, J., et al.Citrus sorting by identification
of the most common defects using multispectral computer vision[J].Journal of
Food Engineering,2007,83:384-393.).Blasc et al. has invented one with reference to ultraviolet induced fluorescence image-forming principle
Set can be used for detecting system (Blanc, P.G.R., Blasco, J., Molt ó, E., the et al.System that citrus rots to detect
for the automatic selective separation of rotten citrus fruit.United States
Patent US2010/0121484A1.2010.), still, ultraviolet easily causes injury to operating personnel, goes out from security standpoint
Hair, UV detection method need to be further improved.Compared with conventional imaging techniques, advanced multispectral imaging is another
The detection technique that can be selected.Aleixos et al. develops the multispectral detecting system of a set of citrus fruit using 2 CCD, colored
CCD provides assessment of the RGB information for fruit size, color, shape of fruit, and monochromatic CCD provides the Near Infrared Information of fruit
For defects detection, but it can not be detected (Aleixos, N., Blasco, J., Navarr to early stage fungal infection fruit
ón,F.,et al.Multispectral inspection of citrus in real-time using machine
vision and digital signal processors[J].Computers and Electronics in
Agriculture,2002,33:121-137.).In newest research, Lorente et al. uses multiplex physicotherapy laser in 2013
The incipient decay citrus as caused by technology for detection Penicillium digitatum infection is scattering into, is found in Visible-to-Near InfaRed
5 wavelength simultaneous Detection results of SPECTRAL REGION preferably (Lorente, D., Zude, M., Regen, C.Early decay
detection in citrus fruit using laser-light backscattering imaging[J]
.Postharvest Biology and Technology,2013,86:424-430.), however, one side this method is being analyzed
During the two-dimensional signal of image is reduced to one-dimensional profile, a large amount of spatial informations will be lost, on the other hand, due to using
Point-like laser light source, must manually make during experiment the regional alignment camera that rots so that light-source angle with
Area-of-interest just matches, and this is not suitable for on-line automaticization detection.Valencia Agricultural Research Institute of Spain (IVIA)
Blasco research teams are using the Hyperspectral imager based on adjustable liquid crystal filter to incipient decay caused by Penicillium digitatum
Citrus is identified ([1] G ó mez-Sanchis, J., Mart í n-Guerrero, J.D., Soria-Olivas, E., et
al.Detecting rottenness caused by Penicillium genus fungi in citrus fruits
using machine learning techniques[J].Expert Systems with Applications,2012,
39:780-785；[2]Gómez-Sanchis,J.,Blasco,J.,Soria-Olivas,E.,et al.Hyperspectral
LCTF-based system for classification of decay in mandarins caused by
Penicillium digitatum and Penicillium italicum using the most relevant bands
and non-linear classifiers[J].Postharvest Biology and Technology,2013,82:76-
86；[3]Lorente,D.,Blasco,J.,Serrano A.J.,et al.Comparison of ROC Feature
Selection Method for the Detection of Decay in Citrus Fruit Using
Hyperspectral Images[J].Food and bioprocess technology,2013,6:3613-3619.).So
And their research is merely focusing on the extraction of visible-near-infrared spectrum complex characteristic, the spatial information of image is have ignored to sense
The feature representation in interest region, be not suitable for the automatic detection of Penicillium digitatum infection fruit.
The content of the invention
The present invention provides one kind rotten citrus image detecting method as caused by infecting Penicillium digitatum, can be by citrus surface
The mould infected zone for being difficult to detect carries out clear visible and shown, and provides effective image processing and analyzing method.
The technical scheme for realizing above-mentioned purpose of the present invention is：
It is a kind of that rotten citrus image detecting method is caused by mould infection, it is characterised in that to comprise the following steps：
A：Gather the Single wavelength spectrum picture of 4 characteristic wave strong points of citrus fruit to be measured, 4 characteristic wavelength λ1, λ2,
λ3And λ4Respectively 575nm, 698nm, 810nm and 969nm；
B：Fruit surface brightness disproportionation correction is carried out to 4 characteristic wavelength images；
C：4 characteristic wavelength λ after step B corrections1, λ2, λ3And λ4Imaging importing obtain characteristic wavelength combination image I：
Stacked system is characteristic wavelength image and the corresponding loading coefficient sum of products；
D：The characteristic wavelength combination image I that step C is obtained is converted into thumbnail F1；
E：Extract R component (Red) image of thumbnail F1 three primary colours components and image preprocessing is carried out to it；
F：Thumbnail F2 is again converted to the step E images obtained and extracts the R of thumbnail three primary colours component
Component image；
G：Target area segmentation is carried out to the R component image obtained in step F and the target area to being split is justified
Shape degree judges.
Wherein, the method for step A collections citrus fruit characteristic wavelength to be measured is：
1) extract fungal infection sample rot region and normal fruit sample normal tissue regions wave-length coverage 500~
Visible-near-infrared spectrum in 1050nm, form spectra collection；
2) characteristic spectrum is weighted, then carrying out principal component to rotten region and normal tissue regions characteristic spectrum gathers
Alanysis；
3) loading coefficient (weight coefficient) curve map of first principal component all wavelengths point, abscissa table in curve map are obtained
Oscillography is grown, and ordinate represents that each wavelength corresponds to load value；
4) wavelength of abscissa corresponding to curve peak and valley is characterized wavelength, and the image of characteristic wave strong point is as characteristic pattern
Picture, load value corresponding to characteristic wavelength image are corresponding loading coefficient.
Step A number of samples can be 10~100 fungal infection samples and 10~100 depending on actual conditions
Individual normal fruit sample；
Specifically, the fungal infection sample is the citrus of Penicillium digitatum infection, is taken in step 4) maximum three on curve
Individual peak and a paddy, characteristic wavelength λ1, λ2, λ3And λ4For 575nm, 698nm, 810nm and 969nm, corresponding loading coefficient is
0.054,0.0469,0.04417 and 0.01436.
Then the formula of step C stacked systems is：
I=0.054 λ1+0.0469λ2+0.04417λ3+0.01436λ4 (1)
Wherein, the step of brightness disproportionation bearing calibration described in step B is as follows：
1) 4 characteristic wavelength image additions are obtained and image Asum；
And image A 2)sumDivided by 4 obtain average image Amean；
3) image after each single band brightness of image correction is obtained using following formula：
Ai_correct=Ai/Amean (2)
In formula, AiExpression is located at characteristic wavelength λ1, λ2, λ3And λ4The characteristic wavelength image at place, Ai_correctAfter expression is corrected
Characteristic wavelength image.
Wherein, it is pseudocolor transformation side characteristic wavelength combination image I to be converted into the implementation method of thumbnail in step D
Method, one kind in gray scale top and bottom process, spatial domain gray level color transformation method or two kinds.
Further, the pseudocolor transformation method need to do twice gray level image to the conversion of thumbnail, selected index
Image conversion method is gray scale top and bottom process, and it is 16 layers to 256 layers that former ash degree is image layered during each thumbnail conversion.
Preferably, the R component image of step E extractions thumbnail three primary colours component is filtered denoising and gray scale lifting is pre-
Processing.
Wherein, the transform method for being again converted to thumbnail F2 is gray scale top and bottom process, and former ash degree is image layered during conversion
For 16 layers to 256 layers.
Wherein, the step G is split using global pre-value to region of being rotted caused by Penicillium digitatum infection, after segmentation
Binaryzation defect area need to carry out circularity judgement, judgment threshold 0.85.Circularity calculation formula:
R=4 π A/P2 (3)
R represents circularity, and A represents marked region area, and P represents marked region girth.
The beneficial effects of the present invention are：
Method proposed by the present invention, by gathering the Single wavelength spectrum picture of 4 characteristic wave strong points of citrus fruit to be measured, knot
Closing image combinatorial formula and image processing method realizes citrus fungal infection region inconspicuous and clearly visualizes display, from
And the citrus that rotted caused by mould infection is identified more effectively, exactly, the present invention relates to less characteristic wavelength, image
Processing Algorithm is easy and effective, has larger application prospect in the detection of online Quality Parameters in Orange.The method of the present invention is high to research and development
End citrus fruit integrated quality classification is equipped, reduction citrus loses after adopting, it is significant to increase farmers' income
Brief description of the drawings
Fig. 1 is the inoculation metainfective fruit photo of Penicillium digitatum.
Fig. 2 is that mycotic infection fruit area-of-interest (filled circles region) spectrum extracts schematic diagram.
Fig. 3 is characterized spectrum classified analysis on major constituents figure.
Fig. 4 is weight (loading) curve map of each spectrum point obtained based on first principal component.
Fig. 5 is the 4 characteristic wavelength single band images obtained based on principal component analysis.
Fig. 6 is characterized Luminance Distribution analysis chart before wavelength image gamma correction.
Fig. 7 is characterized Luminance Distribution analysis chart after wavelength image gamma correction.
Fig. 8 is characterized wavelength combination image.
Fig. 9 is the thumbnail obtained after combination image is changed.
Figure 10 is the R component image of extraction Fig. 9 thumbnail three primary colours components.
Figure 11 is the thumbnail obtained after second of gray level image converts.
Figure 12 is that the R component image based on Figure 11 thumbnail three primary colours components carries out target area segmentation threshold selection.
Figure 13 is Threshold segmentation result figure.
Figure 14 is second of thumbnail of special case.
Figure 15 is second of thumbnail R component image of special case.
Figure 16 is special case Threshold segmentation result figure.
Embodiment
The present invention is now illustrated with following most preferred embodiment, but is not limited to the scope of the present invention.
Unless otherwise instructed, used method is this area conventional technical means to name in embodiment.
Embodiment 1：
The present invention provide it is a kind of rotten citrus image detecting method is caused by mould infection, can will be difficult to the sense that detects
Dye region carries out visible and simultaneously provides effective image processing and analyzing method, before the inventive method implementation, first has to pair
Normal fruit area-of-interest component spectra and mycotic infection fruit area-of-interest component spectra carry out classified analysis on major constituents.Prepare
120 citrus fruit oranges, 120 samples include normal fruit and each 60 samples of mycotic infection fruit, wherein mycotic infection fruit
Obtained using Inoculation Method, specific method is：Infection Penicillium digitatum is cut at the citrusfruit morbidity of natural occurrence
Penicillium digitatum orange peel tissues, 1min is sterilized with 75% medicinal alcohol, spore under aseptic water washing,
Spore is dissolved in sterilized water and forms spores solution, spores solution is then inoculated with by normal fruit using syringe, inoculation depth is big
About 10mm under orange peel.3 are then placed in plastic culture case (environment temperature 25-27 degree, relative humidity 96%~98%)
My god, form a diameter of 10~15mm approaches circular infected zone, and now infected zone is with the naked eye relatively difficult to, Fig. 1 institutes
Be shown as by be inoculated with the metainfective fruit of Penicillium digitatum (artwork for colour, fungal infection region is slightly deep yellow, with citrus fruit
Skin is without obvious control).
Using Visible-to-Near InfaRed Hyperspectral imager (ImSpector V10E, Spectral Imaging Ltd,
Oulu, Finland) obtain sample high spectrum image, it is contemplated that and noise, wave-length coverage are 500~1050nm to image spectrum end to end
High spectrum image is used to analyze, and obtaining 120 panel height spectrum pictures altogether includes normal fruit and each 60 samples of mycotic infection fruit, then
Normal fruit area-of-interest and mycotic infection fruit area-of-interest characteristic spectrum are extracted, area-of-interest is circular, pixel 100
Individual pixel, it is illustrated in figure 2 mycotic infection fruit area-of-interest (filled circles region) spectrum extraction schematic diagram, each sample extraction
One characteristic spectrum forms spectrum set, and therefore, final spectrum set is including 60 rotten Region feature spectrums and 60 just
Chang Guo tissue signatures spectrum.Then two class spectrum are all carried out with twice of weighting enhancing, and principal component is carried out to the spectrum after weighting
Cluster analysis, as shown in figure 3, as can be seen from the figure first principal component can have to two class area-of-interest tissue spectrums
Effect is distinguished.Fig. 4 represents weight (loading) curve map of each spectrum point obtained based on first principal component, horizontal in curve map
Coordinate representation wavelength, ordinate represent that each wavelength corresponds to load value, and wherein curvilinear characteristic peak and valley is corresponded at abscissa wavelength
For image as characteristic wavelength image, load value corresponding to characteristic wavelength image is corresponding loading coefficient, and 4 wavelength points include
3 crests (575nm, 698nm and 810nm) and 1 larger trough (969nm) are selected as characteristic wavelength, 4 wavelength pair
The loading values answered are respectively 0.054,0.0469,0.04417 and 0.01436.
Embodiment 2
The characteristic wavelength obtained based on embodiment 1, method of the invention are comprised the following steps：
A：Using Visible-to-Near InfaRed Hyperspectral imager (ImSpector V10E, Spectral Imaging Ltd,
Oulu, Finland) obtain the Single wavelength spectrum picture for gathering 4 characteristic wave strong points of citrus fruit to be measured, 4 characteristic wavelengths point
Not Wei 575nm, 698nm, 810nm and 969nm, as shown in figure 5, it can be seen that the brightness of one side fruit surface very
Uneven, another aspect infected zone compares very unobvious with normal pericarp region, therefore, it is impossible to directly use Single wavelength
Image carries out the extraction of infected zone.
B：Brightness disproportionation correction is carried out to 4 characteristic wavelength images, brightness disproportionation bearing calibration is as follows：
1) 4 characteristic wavelength image additions are obtained and image Asum；
And image A 2)sumDivided by 4 obtain average image Amean；
3) image after each single band brightness of image correction is obtained using following formula：
Ai_correct=Ai/Amean (2)
In formula, AiRepresent the characteristic wavelength image at wavelength 575nm, 698nm, 810nm and 969nm, Ai_correctTable
Show the characteristic wavelength image after being corrected；
Fig. 6 and Fig. 7 represents to carry out illustrating figure using the typical characteristic wavelength image of a width (698nm Single wavelengths image)
Effect before and after the uneven correction of image brightness, Fig. 6 (a), Fig. 6 (b) and Fig. 6 (c) represent image and Fig. 6 before brightness of image correction respectively
(a) on horizontal section line and longitudinal profile line each pixel brightness value, Fig. 7 (a), Fig. 7 (b) and Fig. 7 (c) represent image respectively
In result images and result images after gamma correction on horizontal section line and longitudinal profile line each pixel brightness value, from Fig. 6
(b) and Fig. 6 (c) it can be seen that image before not correcting, either horizontal section line or longitudinal profile line pixel in image
For grey value profile all in middle high both ends bottom, curve is integrally hemispherical, just reflects the Curvature varying of fruit its own face,
This uneven Luminance Distribution will have a strong impact on the analysis and judgement of final result, from Fig. 7 (b) and Fig. 7 (c) it can be seen that figure
As after gamma correction, image either horizontal section line or longitudinal profile line grey scale pixel value are more uniformly spread, effectively
Inhibit spherical fruit surface brightness disproportionation may caused by by mistake segmentation influence.
C：Characteristic wavelength combination image I is calculated according to following formula：
I=0.054 λ1+0.0469λ2+0.04417λ3+0.01436λ4 (1)
In formula：λ1, λ2, λ3And λ4The brightness at characteristic wavelength 575nm, 698nm, 810nm and 969nm is represented respectively
Single wavelength image after correction, to characteristic wavelength image corresponding to coefficient be corresponding Single wavelength image loading coefficient.Fig. 8 is to obtain
The characteristic wavelength combination image obtained, combination image overall brightness is higher, Penicillium digitatum infected zone and normal pericarp regional correlation
Spend unobvious.
D：By characteristic wavelength combination image I that step C is obtained using gray scale top and bottom process by its gray scale be divided into 256 layers it is (not low
In 16 layers) and thumbnail F1 progress contrast visual enhancements are converted to, after showing conversion such as Fig. 9 (artwork is colour)
Thumbnail F1, the combination image I of compares figure 8, it can be found that thumbnail effectively enhances mycotic infection region and normal fruit
The vision control in dermatotome domain.
E：Extract thumbnail F1 three primary colours components R component image, as shown in Figure 10, and it is filtered denoising and
Image overall intensity lifting processing (being herein image overall intensity increase by 100).
F：Gray scale top and bottom process is reused to the pretreated images of step E its gray scale is divided into 256 layers (not less than 16
Layer) and thumbnail F2 is converted to, intensity contrast enhancing further is carried out to mycotic infection region and normal pericarp region, obtained
Thumbnail F2 as shown in figure 11 (artwork for colour), comparison diagram 11 and Fig. 9 are it can be found that what the conversion of the second secondary index obtained
Its fungal infection region of image compares with normal pericarp region to be become apparent, and overall fruit surface noise is small, then, extraction
(mycotic infection region and normal pericarp regional correlation be most in the component image for the R component image of thumbnail three primary colours component
Greatly), as shown in Figure 12 (a)；
G：Target area segmentation is carried out using global single threshold to the R component image obtained in step F, Figure 12 (b) is figure
Pixel grey scale intensity map in 12 (a) section line, as can be seen from the figure target infection area grayscale value be significantly larger than normal fruit
Dermatotome domain, 30 samples are respectively chosen from 60 mycotic infection fruits and 60 normal fruit samples respectively and are used for this hair as training set
The selection of bright threshold value, by the way that 30 in training set normal fruits and 30 fungal infection fruits are handled with the assessment of image, final fungi sense
Dye region segmentation threshold value is set as 175, and the denoising of morphology opening operation and hole filling processing, knot are performed after single threshold separation is implemented
(for the more preferable position for obtaining localized infection region, fruit edge is also shown in result binary map to fruit bianry image as shown in figure 13
As in, the edge is not present in real image handles computing), it can be observed from fig. 13 that mycotic infection region is had
Effect separates.Because infected zone is in generally similar round, in order to further improve the discrimination of fungal infection, in the present invention
Middle final step need to carry out zone marker to the binary image obtained, calculate circularity (the R=4 π of each marked area
A/P2, R expression circularities, A expression marked region areas, P expression marked regions girth), by training set result point
Analysis, circularity judgment threshold are set as 0.85.Judged result is shown in Table 1.
Embodiment 3：Infected zone is located at the sample at fruit edge
As special case, when infected zone is located at fruit edge, it is the most now to carry out image deflects segmentation detection
It is difficult, according to the identical method of embodiment 2, testing result is as shown in figure 16, and Figure 14 (artwork is colour) and Figure 15 distinguish
Second of thumbnail and the R component of the thumbnail are represented, side provided by the invention is can be seen that from Figure 16 testing result
Method, which will not be infected position of the region in fruit image, to be influenceed, even if infected zone is located at fruit most edge in image,
Also gratifying result can be obtained.
The Samples Estimates result of table 1
Using above step, collection sample is assessed to 60 training set samples (normal fruit and each 30 of fungal infection fruit) and 60
This (normal fruit and each 30 of fungal infection fruit) recognition result is as shown in table 1.In terms of embodiment and testing result, the present invention utilizes
Characteristic wavelength and corresponding image processing algorithm obtain extraordinary Penicillium digitatum infection fruit identification, can be used for the type
The rapid automatized detection of fruit, greatly reduces after citrus fruit is adopted and loses.
Embodiment above is only that the preferred embodiment of the present invention is described, and not the scope of the present invention is entered
Row limits, on the premise of design spirit of the present invention is not departed from, technical side of this area ordinary skill technical staff to the present invention
The all variations and modifications that case is made, it all should fall into the protection domain of claims of the present invention determination.
Claims (9)
1. a kind of cause rotten citrus image detecting method by mould infection, it is characterised in that comprises the following steps：
A：Gather the Single wavelength spectrum picture of 4 characteristic wave strong points of citrus fruit to be measured, 4 characteristic wavelength λ1, λ2, λ3And λ4
Respectively 575nm, 698nm, 810nm and 969nm；
B：Fruit surface brightness disproportionation correction is carried out to 4 characteristic wavelength images；
C：4 characteristic wavelength λ after step B corrections1, λ2, λ3And λ4Imaging importing obtain characteristic wavelength combination image I：Superposition
Mode is characteristic wavelength image and the corresponding loading coefficient sum of products；
D：The characteristic wavelength combination image I that step C is obtained is converted into thumbnail F1；
E：Extract the R component image of thumbnail F1 three primary colours components and image preprocessing is carried out to it；
F：Thumbnail F2 is again converted to the step E images obtained and extracts the R component of thumbnail three primary colours component
Image；
G：Target area segmentation is carried out to the R component image obtained in step F and the target area to being split carries out circularity
Judge.
2. citrus image detecting method as claimed in claim 1, it is characterised in that step A gathers citrus fruit feature to be measured
The method of wavelength is：
Step 1) extraction fungal infection sample rot region and normal fruit sample normal tissue regions wave-length coverage 500~
Visible-near-infrared spectrum in 1050nm, form spectra collection；
Step 2) is weighted to characteristic spectrum, and then carrying out principal component to rotten region and normal tissue regions characteristic spectrum gathers
Alanysis；
Step 3) obtains the loading coefficient curve map of first principal component all wavelengths point, and abscissa represents wavelength in curve map, indulges
Each wavelength of coordinate representation corresponds to load value；
The wavelength of abscissa is characterized wavelength corresponding to step 4) curve peak and valley, and the image of characteristic wave strong point is as characteristic pattern
Picture, load value corresponding to characteristic wavelength image are corresponding loading coefficient.
3. citrus image detecting method as claimed in claim 2, it is characterised in that the fungal infection sample is Penicillium digitatum
The citrus of infection, three peaks maximum on curve and a paddy, characteristic wavelength λ are taken in step 4)1, λ2, λ3And λ4For 575nm,
698nm, 810nm and 969nm, corresponding loading coefficient are 0.054,0.0469,0.04417 and 0.01436.
4. citrus image detecting method as claimed in claim 1, it is characterised in that the brightness disproportionation correction side described in step B
The step of method, is as follows：
1) 4 characteristic wavelength image additions are obtained and image Asum；
And image A 2)sumDivided by 4 obtain average image Amean；
3) image after each single band brightness of image correction is obtained using following formula：
Ai_correct=Ai/Amean (2)
In formula, AiExpression is located at characteristic wavelength λ1, λ2, λ3And λ4The characteristic wavelength image at place, Ai_correctRepresent the spy after being corrected
Levy wavelength image.
5. citrus image detecting method as claimed in claim 1, it is characterised in that by characteristic wavelength combination image I in step D
The implementation method for being converted to thumbnail is pseudocolor transformation method, selected from gray scale top and bottom process, spatial domain gray level color transformation
One kind in method or two kinds.
6. citrus image detecting method as claimed in claim 5, it is characterised in that the pseudocolor transformation method need to be done twice
Gray level image is to the conversion of thumbnail, and selected thumbnail transform method is gray scale top and bottom process, and thumbnail conversion every time
When former ash degree it is image layered be 16 layers to 256 layers.
7. the citrus image detecting method as described in claim 1~6 is any, it is characterised in that step E extracts thumbnail three
The R component image of primary color component is filtered denoising and gray scale lifting pretreatment.
8. the citrus image detecting method as described in claim 1~6 is any, it is characterised in that be again converted to thumbnail
F2 transform method is gray scale top and bottom process, and it is 16 layers to 256 layers that former ash degree is image layered during conversion.
9. the citrus image detecting method as described in claim 1~6 is any, it is characterised in that the step G uses global threshold
Value is split to region of being rotted caused by Penicillium digitatum infection, and the binaryzation defect area after segmentation need to carry out circularity and sentence
It is disconnected, judgment threshold 0.85.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510714053.4A CN105427274B (en) | 2015-10-28 | 2015-10-28 | It is a kind of that rotten citrus image detecting method is caused by mould infection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510714053.4A CN105427274B (en) | 2015-10-28 | 2015-10-28 | It is a kind of that rotten citrus image detecting method is caused by mould infection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105427274A CN105427274A (en) | 2016-03-23 |
| CN105427274B true CN105427274B (en) | 2017-12-26 |
Family
ID=55505455
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510714053.4A Active CN105427274B (en) | 2015-10-28 | 2015-10-28 | It is a kind of that rotten citrus image detecting method is caused by mould infection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105427274B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975966A (en) * | 2016-04-21 | 2016-09-28 | 南京农业大学 | Rice grain mildew nondestructive test method |
| CN106442561B (en) * | 2016-08-29 | 2023-10-31 | 北京农业智能装备技术研究中心 | Online image acquisition system and method for detecting citrus peel defects |
| CN107234074A (en) * | 2017-07-19 | 2017-10-10 | 合肥泰禾光电科技股份有限公司 | A kind of timesharing LED double infrared online material sorting devices and method for separating |
| CN110390664A (en) * | 2018-11-30 | 2019-10-29 | 武汉滨湖电子有限责任公司 | One kind being based on the recognition methods of holes filling pavement crack |
| CN109682821B (en) * | 2018-12-04 | 2022-01-11 | 西南大学 | Citrus surface defect detection method based on multi-scale Gaussian function |
| CN109709103B (en) * | 2019-01-17 | 2020-08-11 | 北京农业智能装备技术研究中心 | Orange early-stage rotting identification system and method based on annular stripe polishing imaging |
| CN111060471B (en) * | 2019-11-28 | 2022-09-16 | 北京农业信息技术研究中心 | Transmission imaging detection method for early rotting of citrus |
| CN112083438B (en) * | 2020-09-17 | 2023-05-16 | 中国科学院空天信息创新研究院 | Indoor mold detection device and method based on hyperspectral laser radar |
| CN112229808A (en) * | 2020-09-21 | 2021-01-15 | 佛山国防科技工业技术成果产业化应用推广中心 | Food microorganism detection device and detection method based on multispectral technology |
| CN117911274B (en) * | 2024-03-19 | 2024-05-28 | 吉林大学 | Thermal imaging technology-based wound infection image optimization method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101984346A (en) * | 2010-10-19 | 2011-03-09 | 浙江大学 | Method of detecting fruit surface defect based on low pass filter |
| CN103439270A (en) * | 2013-09-11 | 2013-12-11 | 南京农业大学 | Device and method for detecting cold damage of peach fruits based on hyperspectral images |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| ES2324324B9 (en) * | 2007-02-27 | 2019-03-12 | Roda Iberica S L U | SYSTEM FOR AUTOMATIC SELECTIVE SEPARATION OF CITRUS AFFECTED BY PODREDUMBRE |
-
2015
- 2015-10-28 CN CN201510714053.4A patent/CN105427274B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101984346A (en) * | 2010-10-19 | 2011-03-09 | 浙江大学 | Method of detecting fruit surface defect based on low pass filter |
| CN103439270A (en) * | 2013-09-11 | 2013-12-11 | 南京农业大学 | Device and method for detecting cold damage of peach fruits based on hyperspectral images |
Non-Patent Citations (2)
| Title |
|---|
| Citrus sorting by identification of the most common defects using multispectral computer vision;J. Blasco et al;《Journal of Food Engineering》;20071231;第83卷(第3期);384-393 * |
| 李江波 等.基于高光谱成像技术检测脐橙溃疡.《农业工程学报》.2010,第26卷(第8期),222-228. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105427274A (en) | 2016-03-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105427274B (en) | It is a kind of that rotten citrus image detecting method is caused by mould infection | |
| Bhargava et al. | Fruits and vegetables quality evaluation using computer vision: A review | |
| Sun et al. | Detection of early decay in peaches by structured-illumination reflectance imaging | |
| Cubero et al. | Automated systems based on machine vision for inspecting citrus fruits from the field to postharvest—a review | |
| Tian et al. | Detection of early decay on citrus using LW-NIR hyperspectral reflectance imaging coupled with two-band ratio and improved watershed segmentation algorithm | |
| Li et al. | Fast detection and visualization of early decay in citrus using Vis-NIR hyperspectral imaging | |
| Blasco et al. | Computer vision detection of peel defects in citrus by means of a region oriented segmentation algorithm | |
| Qin et al. | Detection of citrus canker using hyperspectral reflectance imaging with spectral information divergence | |
| Xiao-bo et al. | In-line detection of apple defects using three color cameras system | |
| Tian et al. | Detection of early decay on citrus using hyperspectral transmittance imaging technology coupled with principal component analysis and improved watershed segmentation algorithms | |
| Blasco et al. | Recognition and classification of external skin damage in citrus fruits using multispectral data and morphological features | |
| Fashi et al. | The relationship between the appearance of pomegranate fruit and color and size of arils based on image processing | |
| CN106442561B (en) | Online image acquisition system and method for detecting citrus peel defects | |
| Hu et al. | Classification and characterization of blueberry mechanical damage with time evolution using reflectance, transmittance and interactance imaging spectroscopy | |
| Sun et al. | Detecting decayed peach using a rotating hyperspectral imaging testbed | |
| Gaston et al. | Visible-near infrared hyperspectral imaging for the identification and discrimination of brown blotch disease on mushroom (Agaricus bisporus) caps | |
| CN103528967B (en) | Overdone indigo fruit fruit identification method based on high spectrum image | |
| Kezhu et al. | Identification of diseases for soybean seeds by computer vision applying BP neural network | |
| Xing et al. | Stem-end/calyx identification on apples using contour analysis in multispectral images | |
| Li et al. | Detection of early decay in navel oranges by structured-illumination reflectance imaging combined with image enhancement and segmentation | |
| Yin et al. | Detection of early bruises on loquat using hyperspectral imaging technology coupled with band ratio and improved Otsu method | |
| Alberto et al. | A computer vision system for early detection of anthracnose in sugar mango (Mangifera indica) based on UV-A illumination | |
| Blasco et al. | Quality evaluation of citrus fruits | |
| Yin et al. | Hyperspectral classification for identifying decayed oranges infected by fungi | |
| Janardhana et al. | Computer aided inspection system for food products using machine vision—a review |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |