| Author | Message | ||
     djdahm |
To add to Howard's comment: The scatter from an object varies with the 4th power of the wavelength for particles that are very small compared to the wavelength. In this case, the scatter in the reverse direcion is the same as in the forward direction. For particles that are large compared to the wavelength, there is much less wavelength dependence in the scatter. At very high absoprtion levels for a large particle, the backscatter is defined by the refractive index of the material, not much by the size of the particle. (The refractive index is also not independent of wavelength.) Then there is the fact that any kind of discontinuity that is a few wavelengths large is a particle edge to the light, so knowing the particle size is not that straight forward. The shape also matters. All in all, it's a problem that seems to need experimenting. Don Dahm | ||
| Tony Davies (Td) |
Hello Rob, Just like to emphasise Ian's message. The most famous (and first) in this line was the the detection of the sex of silkworms inside a cocoon. However the most useful for you is probably J. Near Infrared Spectrosc. 7, 213�221 (1999) Detection of grain weevils inside single wheat kernels by a very near infrared two-wavelength model Christopher Ridgeway and John Chambers Central Science Laboratory, Sand Hutton, York, YO41 1LZ, UK Ian A Cowe Foss Electronic Development (UK) Ltd, Millfield Lane Industrial Estate, Wheldrake, York, YO4 6NA, UK These techniques are thought to work because insects have a very characteristic water spectrum. I think you have a chance of making it work for your problem but it will not be easy! Good luck! Best wishes, Tony | ||
| hlmark |
To augment MPDC's comment: it is well-established that scatter varies inversely with the fourth power of the wavelength. Theoretically, then, the further into the NIR you go, the less the extraneous scatter from the paper should affect the image. On the other hand, we could expect the absorbance from both the paper and the insects to increase. Which one will change more will depend on the spectra of both substances; paper should pretty much look (spectrally) like cellulose, while insects would look mostly like what we can lump together as "protein", depending on their state of dehydration. But as stated, reducing the bandwidth to a single wavelength will also reduce the S/N ratio, which will make it harder to work at high speed. Of course, a "single" wavelength really isn't and you have an extra degree of freedom for finding an optimum set of measurement conditions by using different bandwidths. Also, one way to avoid the heat problems would be to filter the incoming broadband light before it impinges on the sample, rather than afterward. Bottom line, as I see it, is that you might just be able to find a "magic" wavelength that will allow you to do what you want to do. But you shouldn't be too disappointed if you can't, because you have many stringent limitations on what you want to do. Howard \o/ /_\ | ||
| robert pounder (Rpounder) |
I will look further in to this. I think the first hurdle was understanding that what is being proposed has some mileage in testing, although if I understand what I have read correctly speed may be an issue both in processing time and also image capture on a moving object. Does anyone have any experience of alternative inspection techniques that would be worth pursuing in parallel>? Thanks Rob | ||
| MPDC |
Maybe you should try to compare the (NIR) spectroscopic signature of contaminated vs uncontaminated samples. If you can get a full NIR-spectrum in a lab somewhere, it might reveal one or two wavelengths where there is a big difference. You could then mount an interference filter in front of your camera (or in front of your source). Be aware that if you only use 1% of the light, this means you need to increase the source, and you can get heat-related problems. The fact that NIR has greater penetration probably also means it will diffuse less, but I am going on thin ice here. Any opinion from experts? Your existing camera should be sensitive to NIR, but it might have a NIR cut-off filter built in which you need to get rid off. | ||
| Ian Michael (Ianm) |
There have been a few papers on detecting insects in JNIRS. If you are interested, the following search should reveal them: http://www.impublications.com/discus/cgi-nirpubcom/nir_srch.pl?file=all&Range=Any&Format=Standard&Terms=insect You can upload images with a message to this Discussion Forum: see the Formatting section under Documentation in the menu in the left frame of this page, or http://s220179185.websitehome.co.uk/discus/board-formatting.html#images | ||
| robert pounder (Rpounder) |
Ok here is a bit more meat to the problem. I have experimented with various methods of using back and top illumination in the IR, and visible spectrum with off the shelf vision systems. So, when using back illumination the product contents can be clearly inspected using density tools(greyscale). Typically I am using a pixel size of 0.2mm across a 1024x1024 array. However, the product being inspected is packaged within paper hence some diffusion takes place resulting in a clearly defined object with blurred internal features. When using rear illumination, only the dark body section of insect contamination is visible, legs & wings cannot be inspected through the various layers. If people are interested I can e-mail images of insect contamination when viewed in the visible spectrum (with a rule to show size) and when inspected using rear IR illumination. So, my aim is to run the belt at 60m/min which based on the above offers a field of view of circa 200mm. The belt width would be 250mm. The products would be typically spaced at 80mm intervals. I have designed a conveyor which traps the product between a top and bottom belt. Vacuum was used in the prototype (to hold the product flat) but the use of vacuum caused inspection problems owing to using a punched belt hence bright spots in the dressing & packaging. If the insect contamination appeared say just in the packaging or on a film type dressing which transmits the light freely I am confident we can detect this type of contamination. It is most difficult when we have a multi layered product, an exmaple being a film substrate, a fleece pad (smaller than the film placed centrally on the dressing), and silicon release papers covering the dressing. I hope this adds a bit of flesh to the problem and inspires some questions! Thanks Rob. | ||
| MPDC |
If you are talking about on-line inspection well... I think you will get further with NIR light than with visible because NIR has a greater penetration depth. This being said, your problem looks like a very difficult one. How fast does your line move? How wide is it? Do you need 100% inspection or random sampling? If you need to inspect more than 50 cm square per second for something as small as an insect, I am not aware of any off-the-shelf technology which could handle this. I think you would need to work in transmission, and use cameras. This means a powerfull NIR source under the tape and then (a) camera(s) over it. You are anyway looking at a custom-built system here. Please tell us some more, this could become a fun topic! | ||
| robert pounder (Rpounder) |
All, Please excuse my ignorance - I have strayed in to this field accidentally, read a raft of information and am interested to understand if this technology has possibilities for the problem set out below. I am building a machine which is used to inspect finished product (plasters) for contamination. The main problem is throughout our testing of the proposed solution which uses visible light some contamination is not inspectable. As a simple example one of our products is a Polyurethane film with a non woven pad. If insect contamination was present on the pad, due to our inspection techniques the absorbed light is so small we cannot detect the insect (The plaster is inside paper packagin). In short - I have a conveyor belt with plasters being transferred along it - I wish to inspect the product for contamination - Am I barking up the wrong tree with this technology? |