CN103886579B - Abrasive particle chain self-adaptive segmentation method orienting online ferrographic image automatic identification - Google Patents
Abrasive particle chain self-adaptive segmentation method orienting online ferrographic image automatic identification Download PDFInfo
- Publication number
- CN103886579B CN103886579B CN201310675708.2A CN201310675708A CN103886579B CN 103886579 B CN103886579 B CN 103886579B CN 201310675708 A CN201310675708 A CN 201310675708A CN 103886579 B CN103886579 B CN 103886579B
- Authority
- CN
- China
- Prior art keywords
- segmentation
- image
- abrasive particle
- num
- binary
- 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
- 230000011218 segmentation Effects 0.000 title claims abstract description 136
- 239000002245 particle Substances 0.000 title claims abstract description 115
- 238000000034 method Methods 0.000 title claims abstract description 72
- 230000000877 morphologic effect Effects 0.000 claims abstract description 9
- 238000005260 corrosion Methods 0.000 claims description 26
- 230000007797 corrosion Effects 0.000 claims description 24
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims description 20
- 238000005520 cutting process Methods 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 19
- 238000004458 analytical method Methods 0.000 claims description 18
- 239000003550 marker Substances 0.000 claims description 17
- 238000002372 labelling Methods 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 14
- 238000005192 partition Methods 0.000 claims description 13
- 239000008187 granular material Substances 0.000 claims description 11
- 238000005530 etching Methods 0.000 claims description 6
- 229910052742 iron Inorganic materials 0.000 claims description 5
- 230000003595 spectral effect Effects 0.000 claims description 4
- 230000006978 adaptation Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000003628 erosive effect Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 238000004026 adhesive bonding Methods 0.000 claims 1
- 239000008188 pellet Substances 0.000 claims 1
- 238000004321 preservation Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 19
- 238000003709 image segmentation Methods 0.000 abstract description 5
- 238000010191 image analysis Methods 0.000 abstract description 2
- 206010063385 Intellectualisation Diseases 0.000 abstract 1
- 238000012544 monitoring process Methods 0.000 description 10
- 230000008901 benefit Effects 0.000 description 6
- 238000000605 extraction Methods 0.000 description 4
- 238000000926 separation method Methods 0.000 description 4
- 238000005299 abrasion Methods 0.000 description 3
- 230000008021 deposition Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- CWYNVVGOOAEACU-UHFFFAOYSA-N Fe2+ Chemical compound [Fe+2] CWYNVVGOOAEACU-UHFFFAOYSA-N 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006698 induction Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012856 packing Methods 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000010183 spectrum analysis Methods 0.000 description 2
- SUBDBMMJDZJVOS-UHFFFAOYSA-N 5-methoxy-2-{[(4-methoxy-3,5-dimethylpyridin-2-yl)methyl]sulfinyl}-1H-benzimidazole Chemical compound N=1C2=CC(OC)=CC=C2NC=1S(=O)CC1=NC=C(C)C(OC)=C1C SUBDBMMJDZJVOS-UHFFFAOYSA-N 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 238000001157 Fourier transform infrared spectrum Methods 0.000 description 1
- 241000272169 Larus Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 240000008042 Zea mays Species 0.000 description 1
- 235000005824 Zea mays ssp. parviglumis Nutrition 0.000 description 1
- 235000002017 Zea mays subsp mays Nutrition 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 235000005822 corn Nutrition 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000005672 electromagnetic field Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 239000000314 lubricant Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000013528 metallic particle Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000011514 reflex Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
- Finish Polishing, Edge Sharpening, And Grinding By Specific Grinding Devices (AREA)
Abstract
Provided is an abrasive particle chain self-adaptive segmentation method orienting online ferrographic image automatic identification. Step one, a transmitted light image and a reflected light image respectively provided by a ferrographic sensor are respectively preprocessed and converted into a binary image and a graying image; step two, the reflected light image Imgf is adopted to perform coarse segmentation on the basis of gray scale morphology; step three, fine-multi-scale binary morphological segmentation is performed on the binary image after coarse segmentation; a variable scale corrosion-expansion algorithm is adopted on each abrasive particle chain to realize segmentation of large and small abrasive particles so that a binary segmentation line is acquired; and step four, the binary segmentation line is superposed on the original transmitted light image and the reflected light image so that an abrasive particle image after segmentation can be acquired. An online abrasive particle chain image segmentation problem can be solved, and the method can also be applied to abrasive particle chain automatic segmentation in a conventional offline ferrographic image so that the method is significant for realizing intellectualization and automation of a ferrographic image analysis technology.
Description
Technical field
The present invention relates to system state machine monitoring technical field, it is related to the automatic analysis technology of ferrum spectrogram picture, especially relates to
And a kind of abrasive particle chain self-adapting division method towards on-line ferrograph image automatic identification.
Background technology
Debris Analysis are the lubricant mediums and the wear particle situation carrying by the detected machine of analysis, obtain machine
The diagnosis of wear-out failure, prediction, maintenance measures are all played vital acting on by the key technology of state of wear information.
As a kind of important Debris Analysis means, traditional offline ferrum spectrum graphical analyses have been successfully applied to industry dress
Standby wear condition monitoring, defines standardized wear state evaluation system.But this technology has the obvious disadvantage that：1）
The sampling and analysing cycle is long, and efficiency is low；2）The experience dependency of analysis result is strong, and multiplicity is low.With truth maintenance technology
Growth requirement, real-time and automatic monitoring become the developing direction of status monitoring, and traditional off-line analysiss technology is faced with newly
Challenge.
On-line ferrograph technology proposed the thirties in last century, obtained quick development, with on-line sensor over nearly 10 years
Appearance, this traditional Spectral Analysis Technology is extended to real-time monitoring and automatically analyze field.Online Debris Image is by ferrum
The unique advantage of spectral analysis technology is presented on on-line monitoring field, but new technical bottleneck limits this technical advantage：Abrasive particle
Adhesion chaining, cluster substantially increase the difficulty obtaining single abrasive particle characteristic information.Although on-line ferrograph sensor and collection
Method is updated, and does not still find effective abrasive particle chain dividing method, and can only provide macroscopical concentration statistical indicator
（IPCA：Percentage area coverage rate）So as to unique advantage cannot be embodied in many online abrasive particle sensors.Obviously, intelligence
The abrasive particle chain segmentation changed has become as the development bottleneck of on-line ferrograph technology.
Online Debris Analysis are fast-developing Oil Monitoring Technique over nearly 10 years, and various sensors become the master of this technology
Technology to be in progress, can be roughly divided into from principle angle：Electricity, optics, magnetic induction and 4 kinds of image：
1）Electric sensor：" Oil check sensor " based on resistance measurement that Itomi et al. develops（Publication number：
JP2002286697）；The electric capacity abrasive particle enumerator based on micro-channel structure that Murali S is developed based on capacitance principle；Britain
Kittiwake company adopts the Metallic particle Sensor and Total Ferrous Debris of inductance principle exploitation
Sensor.
2）Optical pickocff：The LaserNet Fines wear particle monitoring instrument of USN's research；Allison M.Toms etc.
The FTIR spectrum monitoring device of the online oil liquid monitoring of people's exploitation.
3）Magnetic induction sensor：The MetalScan sensor of Canadian Gastop company.
General character shortcoming using the sensor of above-mentioned 3 kinds of principles is to provide abrasion mechanism information, and ferrum spectrogram seems
The important channel of analysis abrasion mechanism.
Obtain because ferrum spectrogram picture has been used up offline ferrograph device, on-line ferrograph sensor technology becomes new technical side
To.Xi'an Communications University developed the first generation and can provide the sensor of ferrum spectrogram picture online calendar year 2001（The patent No.：
01240347.4）.2005, by the use of composite excitation mode and cmos image sensor, this seminar proposed the second filial generation
Online digital picture iron spectral sensor "（Publication number：CN1673733A）, and the short deposited distance of the third generation has been invented in 2008
Image-type on-line ferrograph sensor（Publication number:CN100365413C）, using electromagnetic field active deposition abrasive particle, can provide respectively
The image of large and small abrasive particle.The online Debris Analysis that are retrieved as of on-line ferrograph image provide technical foundation.
Because using magnetic field deposition principle, abrasive particle inevitably produces chaining, packing phenomenon in ferrum spectrogram picture, such as schemes
Shown in 1, significantly impact the accuracy of follow-up abrasive particle feature analysiss although Debris Image can be eliminated by changing deposition parameter
In packing phenomenon, abrasive particle chaining is but still unavoidable from, and therefore abrasive particle chain segmentation has become as the technology of online Debris Analysis
Bottleneck.Abrasive particle chain is segmented in offline image analyzing iron spectrum and equally exists, and due to having been used up manual analyses, only automatically splits
It is limited to the background segment of single feature abrasive particle, and be seldom related to the report of abrasive particle chain segmentation.It is based on " watershed " Mathematical Morphology
Although method obtains accurate particle partition effect in offline ferrum spectrogram picture, due to needing people in cutting procedure
For setting partitioning parameters, therefore cannot be applied to on-line automatic segmentation.More it is based on mathematical morphology automatic division method
In the application of the segmentation field of adhesion granule, the such as image segmentation of cell and corn, main inclusion：1）Based on rim detection and concave point
The separation method of distribution；2）Separation method based on watershed segmentation；3）Based on the separation method separating Searching point algorithm.Above-mentioned
Various particle image separation algorithms are directed to has class circle property, the granule object of adjoining dimensions, it is dfficult to apply to shape and does not advise
Then, the wide particle partition of edge roughness, distribution of sizes.
The present situation of on-line ferrograph technology shows：The intelligent scissor method of abrasive particle chain has become as on-line ferrograph technology and plays it
The technical bottleneck of Debris Analysis advantage, and existing partitioning algorithm cannot be applied to abrasive particle chain image segmentation.
Content of the invention
For the defect of prior art, the present invention provides the abrasive particle chain self adaptation towards on-line ferrograph image automatic identification to divide
Segmentation method, the abrasive particle chain image partition method in conjunction with gray scale and morphologic information is it is achieved that adhesion abrasive particle in on-line ferrograph image
Segmentation；The present invention not only can solve the segmentation problem of online abrasive particle chain image, applies also for conventional offline ferrum spectrogram picture
In abrasive particle chain automatically split, to realizing, the intellectuality of image analysis technology composed by ferrum and automatization is significant.
In order to achieve the above object, the technical scheme is that:
Towards the abrasive particle chain self-adapting division method of on-line ferrograph image automatic identification, step is as follows：
Step one, the transmitted light images that iron spectral sensor is provided and reflected light image are converted to respectively through pretreatment
Binary image and gray level image.
Step 2, adopt reflected light image ImgfCarry out the coarse segmentation based on gray scale morphology：
Every abrasive particle chain in identification gray level image, and using inner marker and two kinds of control marker character marks of external label
Divisible abrasive particle in note abrasive particle chain, adopts watershed transform to obtain the intensity slicing line of above-mentioned divisible abrasive particle further, and
This cut-off rule is added in corresponding binary image and realizes coarse segmentation；
Step 3, carry out finely multiple dimensioned binary morphology segmentation for the binary image after above-mentioned coarse segmentation：
Corrosion-the expansion algorithm adopting mutative scale for every abrasive particle chain realizes the segmentation to size abrasive particle, obtains binary segmentation line；
Step 4, binary segmentation line is added to original transmitted light images and reflected light image can obtain after segmentation
Debris Image.
Reflected light image gray processing described in step one is by each pixel in image（x,y）Rgb value according to formula
（1）Calculate the gray value obtaining this point, final process result is gray level image：
Formula（1）：
Transmitted light images binaryzation pretreatment described in step one includes gray processing, binaryzation and 3 steps of morphology denoising
Suddenly, described gray processing is with reference to formula（1）：By each pixel in image（x,y）Rgb value according to formula（1）Calculate and obtain
The gray value of this point；Described binaryzation is by setting a threshold value, and passes through formula（2）The pixel value of whole image is divided
Become two parts, thus converting the image into binary image：
Formula（2）：
" opening operation " that morphology denoising is constituted using " corrosion " and " expansion " operator inside mathematical morphology and " close
Computing " carries out morphologic filtering to bianry image, removes unrelated noise.
The coarse segmentation of the gray scale morphology described in step 2 is to be with the abrasive particle chain gray value in reflected light gray level image
Process object, by extracting the method that gray feature carries out particle partition, concrete grammar is：With transmitted light picture binaryzation result
It is reference, using every abrasive particle chain in labelling connected domain method acquisition image as process object, calculate abrasive particle chain image
Shade of gray, draws shade of gray figure；The minimum forcing method of application carries out local minimum fusion, thus obtaining improved minimum
Value region, as required inner marker symbol；The image having done inner marker is carried out respectively with range conversion and watershed becomes
Change, the cut-off rule of acquisition accords with as external label；The inside and outside marker character being obtained is added in gradient amplitude image to former ladder
Degree magnitude image modify, make local minimum area only occur in the position of labelling, that is, with marker character present position be new
Minimum position, simultaneously other local minimum area by blast and will delete；With the gradient amplitude image changed as object
Carry out watershed segmentation and obtain the cut-off rule under gray scale morphology；Obtained cut-off rule is superimposed to original transmitted light images
In pre-processed results, obtain coarse segmentation result.
Described in step 3, multiple dimensioned binary morphology segmentation is realized by analyzing the logical place relation between pixel
The segmentation of adhesion abrasive particle, its analysis object is the Debris Image through primary segmentation, or the adhesion abrasive particle without over-segmentation
Image, binary morphology multi-scale division procedure decomposition is the cutting procedure for various sizes of adhesion granule, for one
Abrasive particle chain to be split, carries out little multi-scale segmentation to size less adhesion granule in abrasive particle chain, respectively through " corrosion ", " condition
Expansion ", " obtaining cut-off rule result " process store obtained cut-off rule to " segmentation result " set；To larger-size
Adhesion granule carries out large scale segmentation, and respectively through " corrosion ", " condition expansion ", " obtaining cut-off rule result " process will be obtained
Cut-off rule store to " segmentation result " set in, the cutting procedure between different scale is independent of each other, segmentation yardstick pass through corruption
Erosion to control with expanding number of times, after the segmentation result under obtaining each yardstick, is superimposed to artwork and can obtain final two
Value segmentation result.
The idiographic flow of multiple dimensioned binary morphology segmentation is：It is used for the variable of control flow firstly the need of initialization
NUMdis,NUMtotal-0Value；Next start multi-scale division circulation after reading in cutting object；In segmentation circulation,
Carry out the segmentation of little yardstick abrasive particle first, by using the less structure factor setting, the morphology " corrosion " of set point number
Operation, obtains " core " of condition expansion, is then based on these " cores " and carries out " condition expansion " operation, set two Expandable strip
Part is：（1）Less than artwork region,（2）Different regions is not in contact with each other；Obtain particle partition line and store by expanding；So
After carry out large scale particle partition, using larger structure factor, the morphological erosion operation of more number of times, obtain condition expansion
" core ", be then based on these " cores " carry out the expansive working with the same terms under little yardstick and finally obtain under this yardstick point
Secant simultaneously stores；Multi-scale division circulation is repeated, through successive ignition, meets stop criterion and stop segmentation, obtain all
Cut-off rule be superimposed to target image and be segmentation result.
Described stop criterion is：In order to meet the self adaptation condition of online picture processing, under some yardstick, dividing
Cut before operation starts, need labelling currently with total connected region quantity NUM in segmentation figure picturetotal-0, and in this yardstick
Under the region quantity NUM that can be disappeared due to first time etching operationdis；After this etching operation, total in statistics Corrosion results
Connected region quantity NUMtotal-1；Then newly-increased region quantity NUM under this cutting operationaddFormula can be passed through（3）Ask
?：
Formula（3）：NUMadd=NUMtotal-1-NUMtotal-0+NUMdis
If NUMadd>0, then explanation has new region to produce, and now stops corrosion, starts condition expansion and obtains cut-off rule
Preserve, this multi-scale segmentation terminates, start the corrosive cycle of next yardstick；If NUMadd=0, then new cut zone is described
Produce, then proceed etching operation, until there being new separated region to produce；If NUMadd<0, illustrate that total region quantity is subtracting
Few, at this time there will be no new region to produce, whole cutting procedure should terminate automatically.
The invention provides abrasive particle chain intellectuality dividing method and segmentation stream in a kind of reliable on-line ferrograph image of precision
Journey, whole flow process includes Image semantic classification, and gray scale morphology is split, 3 key steps of binary morphology multi-scale division.Have
Following beneficial effect：
1st, achieve the automatic segmentation of abrasive particle chain image, be that on-line ferrograph graphical analyses solve technical bottleneck, for follow-up
Abrasive particle feature extraction and abrasion mechanism judge to provide necessary basis, thus it is real that concentration output can only be provided at present to be extended to
Ferrous specturm technique in meaning, has greatly promoted technical advantage in on-line monitoring for this technology；2nd, proposed based on gray scale
The integrated approach of morphology coarse segmentation and mutative scale binaryzation morphology fine segmentation greatly improves the precision of particle partition,
It is that follow-up abrasive particle feature extraction has established important basis with analysis；3rd, gray scale morphology information is introduced the segmentation of abrasive particle chain, change
Become the application situation currently can only being analyzed using transmitted light images it is achieved that the synthesis of transmitted light and reflected light image should
With playing resource and the advantage of on-line ferrograph image technique to greatest extent；4th, by mutative scale binaryzation morphological segment method
Introduce abrasive particle chain to split it is achieved that the adaptivenon-uniform sampling of different scale abrasive particle, solve the problems, such as over-segmentation and less divided, improve
Segmentation precision and intelligence degree.
Brief description
Fig. 1 is split clip and pretreatment explanation；Wherein Fig. 1 a is on-line ferrograph reflected light picture；Fig. 1 b is reflected light figure
Piece gray processing；Fig. 1 c is on-line ferrograph transmitted light picture；Fig. 1 d is transmitted light picture binaryzation.
Fig. 2 is technical solution of the present invention flow chart.
Fig. 3 is gray scale morphology cutting techniques conceptual scheme.
Fig. 4 binary morphology splits the technical schematic diagram of different size abrasive particle.
Fig. 5 is multiple dimensioned binary morphology segmentation flow chart.
Fig. 6 is single abrasive chain gray scale morphology cutting procedure example, and wherein Fig. 6 a is gradient amplitude figure；Fig. 6 b is internal mark
Note result figure；Fig. 6 c is external label result figure；Fig. 6 d is amended gradient amplitude figure；Fig. 6 e is the distribution of intensity slicing line
Figure；Fig. 6 f is gray scale morphology coarse segmentation result.
Fig. 7 is gray scale coarse segmentation result, wherein Fig. 7 a cut-off rule；Fig. 7 b cut-off rule is superimposed to gray-scale maps；Fig. 7 c is segmentation
Line is superimposed to binary map.
Fig. 8 is the binary morphology segmentation example of single abrasive chain, and wherein Fig. 8 a is that yardstick 1 corrodes；Fig. 8 b is that yardstick 1 is swollen
Swollen；Fig. 8 c is yardstick 1 cut-off rule；Fig. 8 d is that yardstick 2 corrodes；Fig. 8 e is that yardstick 2 expands；Fig. 8 f is yardstick 3 cut-off rule；Fig. 8 g is
Yardstick 3 corrodes；Fig. 8 h is that yardstick 3 expands；Fig. 8 i is yardstick 3 cut-off rule；Fig. 8 j is artwork；Fig. 8 k is cut-off rule set；Fig. 8 l
It is that cut-off rule is superimposed to artwork；Fig. 8 m is post processing result；Fig. 8 n is artwork Imgo；Fig. 8 o is in Fig. 7-c（4）Single-stranded rough segmentation
Cut knot；Fig. 8 p is in Fig. 7-c（4）Single-stranded thin segmentation result Img.
Fig. 9 on-line ferrograph image full information splits example automatically.
Preferred forms
Below in conjunction with the accompanying drawings present disclosure is described further.
Algorithm in this enforcement is all realized using business software Matlab.
Towards the abrasive particle chain self-adapting division method of on-line ferrograph image automatic identification, with picture shown in Fig. 1 as material, ginseng
According to Fig. 2, step is as follows：
Step one, the transmitted light images respectively iron spectral sensor being provided（Referring to Fig. 1-c）And reflected light image（Referring to
Fig. 1-a）Be converted to binary image and gray level image respectively through pretreatment.
（1）Gray processing is carried out to reflected light image and obtains gray level image, as shown in Fig. 1-b.
By each pixel in image（x,y）Rgb value according to formula（1）Calculate the gray value obtaining this point, final place
Reason result is gray level image.
Formula（1）：
（2）Transmitted light images are carried out respectively with gray processing, binaryzation, morphology denoising obtain binary image, see Fig. 1-d
Shown.
Described transmitted light images binaryzation pretreatment includes gray processing, binaryzation and 3 steps of morphology denoising, institute
The gray processing stated is with reference to formula（1）：By each pixel in image（x,y）Rgb value according to formula（1）Calculate and obtain this point
Gray value；Described binaryzation is by setting a threshold value, and passes through formula（2）The pixel value of whole image is divided into two
Part, thus convert the image into binary image.
Formula（2）：
Morphology denoising refers to " opening operation " that " corrosion " and " expansion " operator inside using mathematical morphology is constituted
" closed operation " carries out morphologic filtering to bianry image, removes unrelated noise.
Step 2：Coarse segmentation based on gray scale morphology is carried out using reflected light image, with reference to Fig. 3, identifies gray processing figure
Every abrasive particle chain in picture, and using inner marker and two kinds of divisible mills controlling in marker character labelling abrasive particle chains of external label
Grain, adopts watershed transform to obtain the intensity slicing line of above-mentioned divisible abrasive particle further, and this cut-off rule is added to correspondence
Binary image in realize coarse segmentation；It is specially：（1）, with the position of abrasive particle chain in transmitted light binary result as reference, pass through
Connected domain method perception reflex light gray level image（1-b referring to the drawings）In abrasive particle chain, recognition result be 6 abrasive particle chains, point
Other particle partition operation is carried out to every abrasive particle chain, only the cutting procedure of No. 4 abrasive particle chains in Fig. 1-b is said to simplify
Bright；（2）, its gradient amplitude image (referring to the drawings 6-a) is calculated using ' sobel ' operator, in figure white-bright zone domain is gradient amplitude
Minimum region Imgmin；（3）, carry out local minimum fusion using the minimum forcing method of application, thus obtaining improved
Minimum region, removes unrelated minimum region and obtains improved minimum region Imgmin ' as inner marker symbol
Imginner is (with reference to as Fig. 6-b）；（4）, Imgmin ' carried out respectively with range conversion and watershed transform obtains external label
Symbol Imgexter (6-c referring to the drawings）；（5）, inside and outside marker character be superimposed to gradient amplitude image Imggrad modify, make
With Matlab function imimposemin, image Img_exter is modified, make local minimum area occur in the position of labelling
Put, be new minimum position with marker character present position, other local minimum area by blast and will be deleted simultaneously, modification
Gradient amplitude image NImg afterwardsgrad(reference picture 6-d）；（6）, with amended gradient amplitude image NImggradCarry out for object
Watershed segmentation obtains the cut-off rule Segline under gray scale morphologygray(reference picture 6-e）, by cut-off rule SeglinegraySuperposition
Obtain gray scale morphology coarse segmentation result (reference picture 6-f to former gray-scale maps）, so far realize the segmentation of No. 4 abrasive particle chains；（7）, weight
Multiple said process, the gray scale morphology cut-off rule superposition that each bar chain is obtained forms cut-off rule set distribution in the target image
Figure（Reference picture 7-a）, cut-off rule distribution is added in binary picture and target gray image respectively it can be seen that entire image
Segmentation effect (reference picture 7-c and Fig. 7-b).So far achieve the coarse segmentation of target image.
Step 3：Carry out finely multiple dimensioned binary morphology segmentation for the binary image after above-mentioned coarse segmentation：
Corrosion-the expansion algorithm adopting mutative scale for every abrasive particle chain realizes the segmentation to size abrasive particle, obtains binary segmentation line；Two
Value multi-scale morphological cutting procedure can be decomposed into the cutting procedure for various sizes of adhesion granule, such as Fig. 4 middle and upper part
Little multi-scale segmentation and bottom large scale segmentation.For an abrasive particle chain to be split, to the less adhesion of size in abrasive particle chain
Granule carries out little multi-scale segmentation（Fig. 4 top）, respectively through the process such as " corrosion ", " condition expansion ", " obtaining cut-off rule result "
Obtained cut-off rule is stored to " segmentation result " set.Larger-size adhesion granule is carried out with large scale segmentation（Fig. 4
Bottom）, respectively through the process such as " corrosion ", " condition expansion ", " obtain cut-off rule result " by obtained cut-off rule storage
To " segmentation result " set.Cutting procedure between different scale is independent of each other, and segmentation yardstick passes through corrosion and expands number of times
To control.After segmentation result under obtaining each yardstick, it is superimposed to artwork and can obtain final binary segmentation result.Tool
Body is as follows：
With reference to Fig. 8, it is described in detail to process is realized based on the binary morphology multi-scale division of transmitted light：
（1）Little yardstick particle partition prepares：Determine the size of corrosion yardstick and structural element；
Yardstick 1：
（2）Labelling original image Imgo（8-n）In connected region quantity NUMtotal-0(currently for 1), count present image
Middle area is less than connected region quantity NUM of corrosion yardstick 1dis（Current is 0）（Corrosion with yardstick 1 is disappeared by these regions
Lose）；
（3）Treat segmentation figure to corrode yardstick 1 as ImgoCarry out corroding " core " obtaining under segmentation yardstick 1（Fig. 8-a）, sentence
Quantity NUM of connected region after disconnected corrosiontotal-1（Current is 2）；
（4）Carry out end condition judgement（With reference to Fig. 5）, NUMadd=2-1+0=1>0 it is not necessary to terminate；
（5）These " cores " of use condition expansion process obtains the segmentation result (8-b) under yardstick 1；
（6）With the segmentation result under yardstick 1, the cut-off rule also or under computing acquisition yardstick 1 is carried out to bianry image img0
(8-c)；
Yardstick 2：
（7）A multi-scale segmentation result on labelling（8-b）In connected region quantity NUMtotal-0(currently for 2), statistics
In present image, area is less than connected region quantity NUM of corrosion yardstick 2dis（Current is 1）（These regions are by with yardstick 1
Corrode and disappear）；
（8）Treat segmentation figure to corrode yardstick 2 as ImgoCarry out corroding " core " obtaining under segmentation yardstick 2（Fig. 8-d）, sentence
Quantity NUM of connected region after disconnected corrosiontotal-1（Current is 2）；
（9）Carry out end condition judgement（With reference to Fig. 5）, NUMadd=2-2+1=1>0 it is not necessary to terminate；
（10）These " cores " of use condition expansion process obtains the segmentation result (8-e) under yardstick 2；
（11）With the segmentation result under yardstick 2, the cut-off rule also or under computing acquisition yardstick 2 is carried out to bianry image img0
(8-f)；
Yardstick 3：
（12）A multi-scale segmentation result on labelling（8-e）In connected region quantity NUMtotal-0(currently for 2), statistics
In present image, area is less than connected region quantity NUM of corrosion yardstick 3dis（Current is 0）（These regions are by with yardstick 3
Corrode and disappear）；
（13）Treat segmentation figure to corrode yardstick 3 as ImgoCarry out corroding " core " obtaining under segmentation yardstick 3（Fig. 8-g）,
Quantity NUM of connected region after judging to corrodetotal-1（Current is 3）；
（14）Carry out end condition judgement（With reference to Fig. 5）, NUMadd=3-2+0=1>0 it is not necessary to terminate；
（15）These " cores " of use condition expansion process obtains the segmentation result (8-h) under yardstick 3；
（16）With the segmentation result under yardstick 3, the cut-off rule also or under computing acquisition yardstick 3 is carried out to bianry image img0
(8-i)
Yardstick 4：
（17）A multi-scale segmentation result on labelling（8-h）In connected region quantity NUMtotal-0(currently for 3), statistics
In present image, area is less than connected region quantity NUM of corrosion yardstick 4dis（Current is 0）（These regions are by with yardstick 4
Corrode and disappear）；
（18）Treat segmentation figure to corrode yardstick 4 as ImgoCarry out corroding " core " obtaining under segmentation yardstick 4（Fig. 8-j）,
Quantity NUM of connected region after judging to corrodetotal-1（Current is 3）；
（19）Carry out end condition judgement（With reference to Fig. 5）, NUMadd=3-3+0=0, meets end condition, terminates；
（20）By the cut-off rule obtaining under each yardstick superposition（8-k）；
（21）Cut-off rule is superimposed to artwork（8-l）；
（22）Usable floor area threshold method obtains post processing result（8-m）；
So far, binary morphology multi-scale division process terminates.
It is applied to the process description based on gray scale morphology coarse segmentation result for the method below：
（23）Read in the rough segmentation secant that in step 2, gray scale morphology is obtained, and the binary image that is added to（Reference picture
7-c）In.Abrasive particle chain in image is identified by connected domain method, respectively particle partition operation is carried out to every abrasive particle chain, in order to
It is only right to simplify（4）The cutting procedure of number abrasive particle chain illustrates；
（24）, reading in coarse segmentation result taking " 4 " in Fig. 7-c number abrasive particle chain as a example（Fig. 8-o）Afterwards, repeat above-mentioned
（2）~（22）Described multiple dimensioned binary morphology segmentation, exports final segmentation result Img (Fig. 8-p)；
（25）With reference to Fig. 9, realize the segmentation to all of abrasive particle chain, identical operation carried out to each abrasive particle chain,
Step 4, binary segmentation line is added to original transmitted light images and reflected light image can obtain after segmentation
Debris Image.
The segmentation result overlay edge operator that binary morphology is obtained obtains edge（Fig. 9-m）, and it is superimposed to artwork
As final segmentation result design sketch（Fig. 9-n）,
The binary segmentation result that the present invention is finally obtained（Fig. 9-l）For feature extraction；Read in segmentation result two-value
Form（Fig. 9-l）, carry out the extraction of single abrasive characteristic parameter, extract quantity, area, girth and the equivalent dimension of abrasive particle；Specifically
Comprise the following steps：
（1）Extract abrasive particle quantity：Labelling UNICOM region, connected region n of output is abrasive particle quantity；
（Following characteristics all launch for a certain labeled single abrasive particle）
（2）Area：The pixel number that statistics belongs to tag field is the area A in this domain；
（3）Girth：Obtain region contour using boundary operator, follow the trail of the girth C obtaining this domain using chain code；
（4）Equivalent diameter：Area according to abrasive particle and girth, calculate the equivalent dimension of abrasive particle
Wherein, l is the actual size represented by unit pixel；
Final parameter extracts result and stores to data base.
Claims (6)
1. towards on-line ferrograph image automatic identification abrasive particle chain self-adapting division method it is characterised in that step is as follows：
Step one, the transmitted light images that iron spectral sensor is provided and reflected light image are converted to two-value respectively through pretreatment
Change image and gray level image；
Step 2, adopt reflected light image ImgfCarry out the coarse segmentation based on gray scale morphology：
Every abrasive particle chain in identification gray level image, and marker character labelling mill is controlled using inner marker and two kinds of external label
Divisible abrasive particle in pellet chain, further adopt watershed transform obtain above-mentioned divisible abrasive particle intensity slicing line, and by this
Cut-off rule is added in corresponding binary image and realizes coarse segmentation；
Step 3, carry out finely multiple dimensioned binary morphology segmentation for the binary image after above-mentioned coarse segmentation：For
Every abrasive particle chain adopts the corrosion-segmentation to size abrasive particle for expansion algorithm realization of mutative scale, obtains binary segmentation line；
Step 4, binary segmentation line is added to original transmitted light images and reflected light image can obtain the abrasive particle after segmentation
Image.
2. the abrasive particle chain self-adapting division method towards on-line ferrograph image automatic identification according to claim 1, it is special
Levy and be：Reflected light image gray processing described in step one is according to formula by the rgb value of each pixel (x, y) in image
(1) calculate the gray value obtaining this point, final process result is gray level image,
Formula (1)：
Transmitted light images binaryzation pretreatment described in step one includes gray processing, binaryzation and 3 steps of morphology denoising,
Described gray processing is with reference to formula (1)：The rgb value of each pixel (x, y) in image is calculated according to formula (1) and obtains this point
Gray value；Described binaryzation is by setting a threshold value, and by formula (2), the pixel value of whole image is divided into two
Individual part, thus converting the image into binary image,
Formula (2)：
" opening operation " that morphology denoising is constituted using " corrosion " and " expansion " operator inside mathematical morphology and " close fortune
Calculate " bianry image is carried out with morphologic filtering, remove unrelated noise.
3. the abrasive particle chain self-adapting division method towards on-line ferrograph image automatic identification according to claim 1, it is special
Levy and be：The coarse segmentation of the gray scale morphology described in step 2 is with the abrasive particle chain gray value in reflected light gray level image for place
Reason object, by extracting the method that gray feature carries out particle partition, concrete grammar is：With transmitted light picture binaryzation result it is
Reference, using every abrasive particle chain in labelling connected domain method acquisition image as process object, calculates the ash of abrasive particle chain image
Degree gradient, draws shade of gray figure；The minimum forcing method of application carries out local minimum fusion, thus obtaining improved minimum
Region, as required inner marker symbol；Respectively range conversion and watershed transform are carried out to the image having done inner marker,
The cut-off rule obtaining accords with as external label；The inside and outside marker character being obtained is added in gradient amplitude image to former gradient width
Degree image is modified, and makes local minimum area only occur in the position of labelling, is new pole with marker character present position
Little value position, simultaneously other local minimum area by blast and will delete；Carried out with the gradient amplitude image changed for object
Watershed segmentation obtains the cut-off rule under gray scale morphology；Obtained cut-off rule is superimposed to the pre- place of original transmitted light images
In reason result, obtain coarse segmentation result.
4. the abrasive particle chain self-adapting division method towards on-line ferrograph image automatic identification according to claim 1, it is special
Levy and be：Described in step 3, multiple dimensioned binary morphology segmentation is realized gluing by analyzing the logical place relation between pixel
The even segmentation of abrasive particle, its analysis object is the Debris Image through primary segmentation, or the adhesion abrasive particle figure without over-segmentation
Picture, binary morphology multi-scale division procedure decomposition is the cutting procedure for various sizes of adhesion granule, treats for one
Segmentation abrasive particle chain, the adhesion granule little to size in abrasive particle chain carries out little multi-scale segmentation, and respectively through " corrosion ", " condition is swollen
Swollen ", " obtain cut-off rule result " process stores obtained cut-off rule to " segmentation result " set；The adhesion big to size
Granule carries out large scale segmentation, divides obtained respectively through " corrosion ", " condition expansion ", " obtaining cut-off rule result " process
Secant store to " segmentation result " set in, the cutting procedure between different scale is independent of each other, segmentation yardstick pass through corrosion with
Expand number of times controlling, after the segmentation result under obtaining each yardstick, be superimposed to artwork and can obtain final two-value and divide
Cut result.
5. the abrasive particle chain self-adapting division method towards on-line ferrograph image automatic identification according to claim 3, it is special
Levy and be：The idiographic flow of multiple dimensioned binary morphology segmentation is：It is used for the variable of control flow firstly the need of initialization
NUMdis,NUMtotal-0Value, NUMdisIt is the region quantity that first time etching operation can disappear, NUMtotal-0Total in segmentation figure picture
Connected region quantity；Next start multi-scale division circulation after reading in cutting object；In segmentation circulation, first
Carry out the segmentation of little yardstick abrasive particle, by using the less structure factor setting, morphology " corrosion " operation of set point number
Obtain " core " of condition expansion, be then based on these " cores " and carry out " condition expansion " operation, two set exaggerated conditions are：
(1) it is less than artwork region, (2) different region is not in contact with each other；Obtain particle partition line and store by expanding；Then carry out
Large scale particle partition, using larger structure factor, the morphological erosion operation of more number of times, obtains condition expansion
" core ", is then based on these " cores " and carries out the expansive working with the same terms under little yardstick and finally obtain the segmentation under this yardstick
Line simultaneously stores；Multi-scale division circulation is repeated, through successive ignition, meets stop criterion and stop segmentation, obtain all of
Cut-off rule is superimposed to target image and is segmentation result.
6. the abrasive particle chain self-adapting division method towards on-line ferrograph image automatic identification according to claim 4, it is special
Levy and be：In order to meet the self adaptation condition of online picture processing, under some yardstick, before cutting operation starts, need
Want labelling currently with total connected region quantity NUM in segmentation figure picturetotal-0, and due to first time corrosion behaviour under this yardstick
The region quantity NUM that work can disappeardis；After this etching operation, total connected region quantity in statistics Corrosion results
NUMtotal-1；Then newly-increased region quantity NUM under this cutting operationaddTried to achieve by formula (3)：
Formula (3)：NUMadd=NUMtotal-1-NUMtotal-0+NUMdis
If NUMadd>0, then explanation has new region to produce, and now stops corrosion, starts condition expansion and obtains cut-off rule preservation,
This multi-scale segmentation terminates, and starts the corrosive cycle of next yardstick；If NUMadd=0, then illustrate that new cut zone is produced
Raw, then proceed etching operation, until there being new separated region to produce；If NUMadd<0, illustrate that total region quantity is subtracting
Few, at this time there will be no new region to produce, whole cutting procedure should terminate automatically.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310675708.2A CN103886579B (en) | 2013-12-11 | 2013-12-11 | Abrasive particle chain self-adaptive segmentation method orienting online ferrographic image automatic identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310675708.2A CN103886579B (en) | 2013-12-11 | 2013-12-11 | Abrasive particle chain self-adaptive segmentation method orienting online ferrographic image automatic identification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103886579A CN103886579A (en) | 2014-06-25 |
| CN103886579B true CN103886579B (en) | 2017-02-08 |
Family
ID=50955453
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310675708.2A Expired - Fee Related CN103886579B (en) | 2013-12-11 | 2013-12-11 | Abrasive particle chain self-adaptive segmentation method orienting online ferrographic image automatic identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103886579B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104484675B (en) * | 2014-12-15 | 2017-10-31 | 上海海事大学 | A kind of iron spectrum abrasive particle texture feature extraction and the method for pattern-recognition |
| CN106041946B (en) * | 2016-05-23 | 2017-02-22 | 广东工业大学 | Image-processing-based robot polishing production method and production system applying same |
| CN105798927B (en) * | 2016-05-23 | 2017-02-08 | 广东工业大学 | Crucible defect grinding processing method based on image and processing system with crucible defect grinding processing method based on image |
| CN106004203B (en) * | 2016-05-23 | 2017-02-22 | 广东工业大学 | Hole decorating plate machining method based on gray levels and machining system applying hole decorating plate machining method based on gray levels |
| CN107063946B (en) * | 2016-12-23 | 2023-04-28 | 天津市三特电子有限公司 | Video green ball particle diameter instrument and green ball particle diameter state detection method thereof |
| CN107103594B (en) * | 2017-05-18 | 2020-04-14 | 中国工程物理研究院激光聚变研究中心 | Quantitative measurement and characterization method for abrasive wear of micro-powder diamond grinding wheel |
| CN107633485B (en) * | 2017-08-07 | 2022-03-25 | 百度在线网络技术（北京）有限公司 | Face brightness adjusting method, device, equipment and storage medium |
| CN108305259B (en) * | 2018-02-06 | 2020-03-24 | 西安交通大学 | Multi-texture feature fusion type automatic abrasive particle type identification method |
| CN108389216B (en) * | 2018-02-06 | 2020-06-26 | 西安交通大学 | Local self-adaptive threshold segmentation method for online ferrographic image abrasive particle identification |
| CN108446706B (en) * | 2018-02-27 | 2021-01-19 | 西安交通大学 | Automatic abrasive grain material identification method based on principal component extraction of colors |
| CN110276280B (en) * | 2019-06-06 | 2021-06-04 | 刘嘉津 | Optical processing method for automatically identifying crop pest images |
| CN110223299B (en) * | 2019-06-12 | 2021-06-18 | 重庆邮电大学 | Abrasive particle segmentation method based on deposition process |
| CN110473213B (en) * | 2019-08-16 | 2021-08-10 | 重庆邮电大学 | Iron spectrum image segmentation method, device and system based on prior lattice |
| CN112053355B (en) * | 2020-09-16 | 2023-03-10 | 昆明理工大学 | Cell image segmentation method |
| CN113409229B (en) * | 2021-08-20 | 2022-04-08 | 南京航空航天大学 | Method for evaluating contour of abrasive particles of large-abrasive-particle superhard abrasive grinding wheel |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6970606B2 (en) * | 2002-01-16 | 2005-11-29 | Eastman Kodak Company | Automatic image quality evaluation and correction technique for digitized and thresholded document images |
| CN102494976A (en) * | 2011-11-18 | 2012-06-13 | 江苏大学 | Method for automatic measurement and morphological classification statistic of ultra-fine grain steel grains |
-
2013
- 2013-12-11 CN CN201310675708.2A patent/CN103886579B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6970606B2 (en) * | 2002-01-16 | 2005-11-29 | Eastman Kodak Company | Automatic image quality evaluation and correction technique for digitized and thresholded document images |
| CN102494976A (en) * | 2011-11-18 | 2012-06-13 | 江苏大学 | Method for automatic measurement and morphological classification statistic of ultra-fine grain steel grains |
Non-Patent Citations (4)
| Title |
|---|
| Automated classification of wear particles based on their surface texture and shape features;Stachowiak G P et al.;《Tribology International》;20081231;34-43 * |
| 图像可视在线铁谱传感器的图像数字化处理技术;武通海等;《机械工程学报》;20080930;第44卷(第9期);83-87 * |
| 基于变尺度形态学的遥感图像边缘检测算法;焦斌亮等;《光学仪器》;20050831;第27卷(第4期);13-17 * |
| 基于背景减除的彩色铁谱图像自动分割;王静秋等;《润滑与密封》;20110531;第36卷(第5期);48-52 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103886579A (en) | 2014-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103886579B (en) | Abrasive particle chain self-adaptive segmentation method orienting online ferrographic image automatic identification | |
| CN109492581B (en) | Human body action recognition method based on TP-STG frame | |
| CN102005034B (en) | Remote sensing image segmentation method based on region clustering | |
| CN104036239B (en) | Fast high-resolution SAR (synthetic aperture radar) image ship detection method based on feature fusion and clustering | |
| CN101800890B (en) | Multiple vehicle video tracking method in expressway monitoring scene | |
| Jahedsaravani et al. | An image segmentation algorithm for measurement of flotation froth bubble size distributions | |
| CN110135296A (en) | Airfield runway FOD detection method based on convolutional neural networks | |
| CN103049763B (en) | Context-constraint-based target identification method | |
| CN111047551A (en) | Remote sensing image change detection method and system based on U-net improved algorithm | |
| CN108595414B (en) | Soil heavy metal enterprise pollution source identification method based on source-sink space variable reasoning | |
| CN106056118A (en) | Recognition and counting method for cells | |
| CN109766823A (en) | A kind of high-definition remote sensing ship detecting method based on deep layer convolutional neural networks | |
| CN102496034A (en) | High-spatial resolution remote-sensing image bag-of-word classification method based on linear words | |
| CN108417033A (en) | Expressway traffic accident analysis prediction technique based on multi-dimensional factors | |
| CN104732215A (en) | Remote-sensing image coastline extracting method based on information vector machine | |
| CN103942564A (en) | High-resolution remote sensing image scene classifying method based on unsupervised feature learning | |
| CN108875593A (en) | Visible images weather recognition methods based on convolutional neural networks | |
| CN113033714B (en) | Object-oriented full-automatic machine learning method and system for multi-mode multi-granularity remote sensing image | |
| CN101876987A (en) | Overlapped-between-clusters-oriented method for classifying two types of texts | |
| CN112861931B (en) | Multi-level change detection method, system, medium and electronic device based on difference attention neural network | |
| CN104778442A (en) | Automatic segmentation and counting method of retina cell fluorescence microscopic image | |
| CN105243387A (en) | Open-pit mine typical ground object classification method based on UAV image | |
| CN106203496A (en) | Hydrographic curve extracting method based on machine learning | |
| Bhatt et al. | Spectral indices based object oriented classification for change detection using satellite data | |
| CN106202274A (en) | A kind of defective data automatic abstract sorting technique based on Bayesian network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170208 |