A ceramic plate must be inspected for cracks after it has left the oven... The process currently involves "painting" the plate with blue ink, so that cracks and other features that are at a diferent level than what surface of the plate become aparent... I was wondering if this could be done via image processing... Here are some facts: Cracks may be small(about half an inch) or big(several inches). Cracks that have not been painted with the die are dificult to detect for the naked eye. The whole process, from taking the picture to end of inspection should take 3 seconds tops... I will take some pictures and uploadthem on monday if posible, so you can see that im talking about. I dont think a simple edge detect would work... and have been looking the web for some algorithm to show the contour of the piece.... any ideas? This message was sent using the Comp.DSP web interface on www.DSPRelated.com
revealing cracks in ceramic items...
Started by ●October 10, 2005
Reply by ●October 10, 20052005-10-10
"NightHawk" <nightstalker_099@yahoo.com> wrote in message news:SLedndzVIZrxedTenZ2dnUVZ_s6dnZ2d@giganews.com...> > A ceramic plate must be inspected for cracks after it has left the > oven... > The process currently involves "painting" the plate with blue ink, so that > cracks and other features that are at a diferent level than what surface > of > the plate become aparent... > I was wondering if this could be done via image processing... > Here are some facts: > > Cracks may be small(about half an inch) or big(several inches). > Cracks that have not been painted with the die are dificult to detect for > the naked eye. > The whole process, from taking the picture to end of inspection should > take 3 seconds tops... > > I will take some pictures and uploadthem on monday if posible, so you can > see that im talking about. > > I dont think a simple edge detect would work... and have been looking the > web for some algorithm to show the contour of the piece.... any ideas?It depends on the images and I'm not sure if you meant to do it without the ink. With the ink, the cracks may represent high spatial frequencies depending on the rest of the surface being inspected. High spatial frequencies can be detected in a number of ways: - taking a Fourier Transform over a prescribed area and detecting cracks based on high frequency content exceeding some threshold. - computing the contrast or color range of an area of an image and detecting cracks based on the contrast being higher than some threshold. Without the ink maybe there's illumination tricks you might use. Much machine vision work depends on particular illumination. Again, it depends on the images. I hope this points you in some useful directions. Fred
Reply by ●October 10, 20052005-10-10
Hi NH, When there are cracks, do they go all the way to the other side of the plate ? - If yes, then you do not need to paint, just illuminate from the OTHER side of where the camer is. Of course, the surrounding room must be dark... - If no, then maybe consider the technique used in the aero-space industry to inspect for microscopic cracks: create vibrational surface waves, and detect the response with several sensors placed in various locations on the plate surface. The response will be very different between good and bad plates. Gidi NightHawk wrote:> A ceramic plate must be inspected for cracks after it has left the > oven... > The process currently involves "painting" the plate with blue ink, so that > cracks and other features that are at a diferent level than what surface of > the plate become aparent... > I was wondering if this could be done via image processing... > Here are some facts: > > Cracks may be small(about half an inch) or big(several inches). > Cracks that have not been painted with the die are dificult to detect for > the naked eye. > The whole process, from taking the picture to end of inspection should > take 3 seconds tops... > > I will take some pictures and uploadthem on monday if posible, so you can > see that im talking about. > > I dont think a simple edge detect would work... and have been looking the > web for some algorithm to show the contour of the piece.... any ideas? > > > This message was sent using the Comp.DSP web interface on > www.DSPRelated.com
Reply by ●October 10, 20052005-10-10
"NightHawk" <nightstalker_099@yahoo.com> wrote in message news:SLedndzVIZrxedTenZ2dnUVZ_s6dnZ2d@giganews.com...><snip>> I dont think a simple edge detect would work... and have been looking the > web for some algorithm to show the contour of the piece.... any ideas? >you could try http://www.sciencedirect.com/science?_ob=JournalURL&_cdi=5665&_auth=y&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=d9da1c2ab07f3477fe104fe9f4dd9e33 I think there is also an electrostatic technique for enhancing the visibility, which relies on enhanced fields near fine cracks and other sharp irregularities but I'm not sure how you would find out about this, nor how suitable it would be. Best of Luck - Mike
Reply by ●October 10, 20052005-10-10
Yes, I mean to do it wothout the ink. Ill try and check out the spatial frequency analisis.>Hi NH, > >When there are cracks, do they go all the way to the other side of the >plate ? > >- If yes, then you do not need to paint, just illuminate from the OTHER >side of where the camer is. Of course, the surrounding room must be >dark... > >- If no, then maybe consider the technique used in the aero-space >industry to inspect for microscopic cracks: create vibrational surface >waves, and detect the response with several sensors placed in various >locations on the plate surface. The response will be very different >between good and bad plates. > >Gidi > > >NightHawk wrote: >> A ceramic plate must be inspected for cracks after it has left the >> oven... >> The process currently involves "painting" the plate with blue ink, sothat>> cracks and other features that are at a diferent level than whatsurface of>> the plate become aparent... >> I was wondering if this could be done via image processing... >> Here are some facts: >> >> Cracks may be small(about half an inch) or big(several inches). >> Cracks that have not been painted with the die are dificult to detectfor>> the naked eye. >> The whole process, from taking the picture to end of inspection should >> take 3 seconds tops... >> >> I will take some pictures and uploadthem on monday if posible, so youcan>> see that im talking about. >> >> I dont think a simple edge detect would work... and have been lookingthe>> web for some algorithm to show the contour of the piece.... any ideas? >> >> >> This message was sent using the Comp.DSP web interface on >> www.DSPRelated.com > >This message was sent using the Comp.DSP web interface on www.DSPRelated.com
Reply by ●October 11, 20052005-10-11
Hi NH, After a 2nd thought, there is a simpler low-tech solution (for the case that cracks do not go through to the other side): Do not use paint. Paint has several disadvantages: - it's difficult to remove. - you waste money when painting each tile. - the paint gets into the cracks, thereby lowering the imaging contrast between Therefore, use powder or other small granules: - After pouring the granules/powder on the surface, the plates are delicately shaken/vibrated, to remove the excess from the correct (good) surface, so that granules/powder remains only inside cracks (if there are any) !!!!! - After capturing an image, all granules/powder may be easily removed by a vacuum-cleaner, and re-used on the nextplate !! - The diameter of the granules should be at least 10 times smaller than the average width of cracks. (Eg., if avg width is 1mm, use granules/powder with diameter smaller than 0.1mm) - The color of the granules/powder should be such that will result in the largest possible contrast with the imaging equipment you use. For example, use white powder which reflects low-UV light, on "dull" plates, with a B&W camera. Or use yellow powder on purple plates, and use an appropriate color filter before a B&W cammera. Etc... you get the idea. The image processing should be VERY SIMPLE due to the large contrast: - due to the contrast, there should be a clear probability distribution in each of the good/bad plates: bi-modal: on the good plates, the percentage of "white" should be below 10% (for example), and above 60% in small regions with cracks !! Therefore: - you can perform processing to work on binary images by placing a threshold on the actual captured image values. - perform averaging on "small" regions. If the average value is above/below some predefined threshold in at least one of the regions, then it means there is a crack in that region, therefore reject the tile. Gidi
Reply by ●October 11, 20052005-10-11
NightHawk wrote:> Yes, I mean to do it wothout the ink. Ill try and check out the spatial > frequency analisis.... Vibrational analysis is done by exciting the plate and detecting its vibrations. I can imagine two ways to do the detecting: microphones and Chladni patterns. Any given plate has many modes of vibration,depending on frequency. Since your plates will vary somewhat in thickness, a slightly tunable oscillator might be needed for excitation. Modes are cataloged at http://astronomy.swin.edu.au/~pbourke/modelling/chladni/. The important feature is that a cracked plate won't exhibit a standard pattern, but an intact plate will. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●October 11, 20052005-10-11
Jerry Avins wrote:> NightHawk wrote: > >> Yes, I mean to do it wothout the ink. Ill try and check out the spatial >> frequency analisis. > > > ... > > Vibrational analysis is done by exciting the plate and detecting its > vibrations. I can imagine two ways to do the detecting: microphones and > Chladni patterns. Any given plate has many modes of vibration,depending > on frequency. Since your plates will vary somewhat in thickness, a > slightly tunable oscillator might be needed for excitation. Modes are > cataloged at http://astronomy.swin.edu.au/~pbourke/modelling/chladni/. > The important feature is that a cracked plate won't exhibit a standard > pattern, but an intact plate will. > > JerryJerry's response triggered practical and probably impractical ideas. The practical: Jerry's description reminded me of one of the segment of one of the PBS antique related programs. They were describing quick tests for the quality and soundness of china. Tapping with finger nail. High quality and defect free pieces would ring with a pure distinctive tone. The idea, which is actually more of a question, is "are there other techniques used by "experts" for quick field tests of purported high quality antique ware?" The *less* practical perhaps. [I'd be interested if anyone can shoot holes in idea itself, rather than its possible economic and engineering feasibility aspects] Plate comes off production line. I comes to rest in test jig which assures it is in 'same' position as all other samples if it is not symmetric on every axis. The top is illuminated by laser light which has gone thru beam splitter. The result being an observable interference pattern. The bottom of plate is "illuminated" by a sonic impulse - eg short duration high energy arc [no responsibility taken for nearby sensitive electronics ;] Would the before/after impulse response of "good" plate be consistent enough to distinguish them from a "bad" plate? Would that be a "Chladni Plate" *without* the sand/powder?
Reply by ●October 11, 20052005-10-11
Richard Owlett wrote:> Jerry Avins wrote: > >> NightHawk wrote: >> >>> Yes, I mean to do it wothout the ink. Ill try and check out the spatial >>> frequency analisis. >> >> >> >> ... >> >> Vibrational analysis is done by exciting the plate and detecting its >> vibrations. I can imagine two ways to do the detecting: microphones >> and Chladni patterns. Any given plate has many modes of >> vibration,depending on frequency. Since your plates will vary somewhat >> in thickness, a slightly tunable oscillator might be needed for >> excitation. Modes are cataloged at >> http://astronomy.swin.edu.au/~pbourke/modelling/chladni/. The >> important feature is that a cracked plate won't exhibit a standard >> pattern, but an intact plate will. >> >> Jerry > > > Jerry's response triggered practical and probably impractical ideas. > > The practical: > Jerry's description reminded me of one of the segment of one of the PBS > antique related programs. They were describing quick tests for the > quality and soundness of china. Tapping with finger nail. High quality > and defect free pieces would ring with a pure distinctive tone. > > The idea, which is actually more of a question, is "are there other > techniques used by "experts" for quick field tests of purported high > quality antique ware?" > > > The *less* practical perhaps. > [I'd be interested if anyone can shoot holes in idea itself, rather than > its possible economic and engineering feasibility aspects] > > Plate comes off production line. > > I comes to rest in test jig which assures it is in 'same' position as > all other samples if it is not symmetric on every axis. > > The top is illuminated by laser light which has gone thru beam splitter. > The result being an observable interference pattern. > > The bottom of plate is "illuminated" by a sonic impulse - eg short > duration high energy arc [no responsibility taken for nearby sensitive > electronics ;] > > Would the before/after impulse response of "good" plate be consistent > enough to distinguish them from a "bad" plate?Before and after what?> Would that be a "Chladni Plate" *without* the sand/powder?Yes. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●October 11, 20052005-10-11
Jerry Avins wrote:> Richard Owlett wrote: > >> Jerry Avins wrote: >> >>> NightHawk wrote: >>> >>>> Yes, I mean to do it wothout the ink. Ill try and check out the spatial >>>> frequency analisis. >>> >>> >>> >>> >>> ... >>> >>> Vibrational analysis is done by exciting the plate and detecting its >>> vibrations. I can imagine two ways to do the detecting: microphones >>> and Chladni patterns. Any given plate has many modes of >>> vibration,depending on frequency. Since your plates will vary >>> somewhat in thickness, a slightly tunable oscillator might be needed >>> for excitation. Modes are cataloged at >>> http://astronomy.swin.edu.au/~pbourke/modelling/chladni/. The >>> important feature is that a cracked plate won't exhibit a standard >>> pattern, but an intact plate will. >>> >>> Jerry >> >> >> >> Jerry's response triggered practical and probably impractical ideas. >> >> The practical: >> Jerry's description reminded me of one of the segment of one of the >> PBS antique related programs. They were describing quick tests for the >> quality and soundness of china. Tapping with finger nail. High quality >> and defect free pieces would ring with a pure distinctive tone. >> >> The idea, which is actually more of a question, is "are there other >> techniques used by "experts" for quick field tests of purported high >> quality antique ware?" >> >> >> The *less* practical perhaps. >> [I'd be interested if anyone can shoot holes in idea itself, rather >> than its possible economic and engineering feasibility aspects] >> >> Plate comes off production line. >> >> I comes to rest in test jig which assures it is in 'same' position as >> all other samples if it is not symmetric on every axis. >> >> The top is illuminated by laser light which has gone thru beam splitter. >> The result being an observable interference pattern. >> >> The bottom of plate is "illuminated" by a sonic impulse - eg short >> duration high energy arc [no responsibility taken for nearby sensitive >> electronics ;] >> >> Would the before/after impulse response of "good" plate be consistent >> enough to distinguish them from a "bad" plate? > > > Before and after what?the quiescent versus after applied impulse states> >> Would that be a "Chladni Plate" *without* the sand/powder? > > > Yes. > > Jerry
