Question: Question: I want to do citizen bee science

FortunaWolf is asking a question about spectrometry
Follow this topic

by FortunaWolf | April 18, 2016 15:49 | #13000


What I want to do or know

Hello human colleagues, The technical details are that I'd like to collect spectroscopic data on pollen pellets that honey bees bring back to the lab for the purpose of doing species identification in the field. We want beekeepers and citizen scientists to be able to collect pollen from their hives around the world and through the year, take spectrographic data, and mail their pollen to us. Eventually we'll build a database on pollen species and spectrographic fingerprints so that when they go out next year and query the pollen they'll get an instant answer to what species of pollen their bees are working.

So the technical question is, what are the capabilities of the handheld spectrometer? DVDs have about 1400 lines/mm. Would it give me better resolution to use 1000 lines/mm or would the band go off the sides of the field of view and I would crop my data? What's the limiter to the resolution and how could it be improved? I expect to modify it at a minimum by making a custom mount for the pollen so its always positioned in a repeatable method and then using 1 or 2 defined light sources (a UV LED or laser for fluorescent spectrometry and a full spectrum 400-900nm blackbody source). Any suggestions on light sources? has this been done?

Any success connecting external webcams to android/iphone phones? I'd love it if people were able to use an external webcam that's been molested to remove the IR cut filter, and since this is a field study using phones maximizes portability.

Is the analysis software open source? I think I might want to molest that too in order to automate sending the data to the server and getting the results back.

Background story

We are a large public university bee lab and one of the grad students in my lab had an idea to use camera phones to image pollen pellets and use the RGB data to ID the species. Initial results look very promising and its time to take this to the next level for a post-doc study and extension service.



7 Comments

Hi, FortunaWolf - what resolution do you need, is probably the first question. I think the DVD will probably not be the limiting factor but if the basic kit is not sufficient, there are lots of things you can do to tighten up the resolution, but it's work, so perhaps you can start with a goal. One thing you can do is put it in a rigid housing -- there are some being developed here (see comments especially):

https://publiclab.org/notes/viechdokter/04-15-2016/thinking-about-a-more-stable-spectrometer

Are you building on some existing research you've found on pollen spectrometry, or trying something entirely new?

Re: webcams, there are some phones which can connect to webcams, but there are no UVC drivers on Android, so they work via apps (as far as I can tell). I haven't had success with it, though would love to see it happen. Perhaps if you're looking for miniaturization, a Raspberry Pi would be a way to go?

The software is open source, indeed: https://github.com/publiclab/spectral-workbench

And we recently spun out the JavaScript analysis core for independent use: https://github.com/publiclab/spectral-workbench.js (usable in Node.js, though no live-capture yet).

Hope that helps!

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Hi Warren, I don't know what resolution I need yet, TBH. Initial data from the RGB data from cell phone cameras shows that knowing the geographic location and timing of collection lets us ID most species (the RGB data forming a cloud that occasionally partially overlaps with another species). I want to move to spectrographic data to tease that apart, since I think two species which look the same may have different peaks. I want to build my library with the best resolution possible, so I could do $100-200 and send a handful of those out to initial collaborators. Ideally it would be closer to $50 so I can increase the amount of collaborators and even let people buy the part and assist us. In the end I hope that we'd be able to use the $10 spectrometer for the end user who would be querying the library.

Pollen identification from color was last reported on in the literature in the 1970s. Time to update it.

I considered the RPI, but to add a touchscreen, camera module, battery pack, and GPS... I just don't want to go down that route unless I have to. I like the idea of the phone having an app that does it all and includes a GPS and time tag. We could even use the cellular connection to query the library and get results back in the field.

What USB cameras are being used right now? Would it be better to look for a cheap black and white high resolution camera? What resolution is worth bothering with? I'm thinking black and white because it might not have bayer filters on it and that would increase sensitivity of the sensor across the whole spectrum.

Could I write Public Lab into a grant so that you get some grant money to do software development?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


The camera is listed here: https://publiclab.org/wiki/desktop-spectrometry-kit-3-0#Parts+List

I think a B&W one would be great, but I really don't think they're sold anymore (at least not affordable ones). We're currently limiting to 640x480 capture resolution, but trying to remove that limitation: https://github.com/publiclab/spectral-workbench/issues/104

This might also be addressed in the new Node.js library, as part of https://github.com/publiclab/spectral-workbench.js/issues/16

Could I write Public Lab into a grant so that you get some grant money to do software development?

Well, that sounds really awesome, but I have to direct you to Becki Chall who leads our fundraising: https://publiclab.org/wiki/funding-partnerships-with-public-lab

Whether or not you do that, we can always use help to get the word out about our software platform and where folks can get involved to help drive it forward. There's a lot of info at https://publiclab.org/wiki/developers

Thanks!

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


I wonder if you could do raw data capture off of the webcam (that would also help with your autogain issues too). By removing the bayer filter you can convert it into a monochome camera with higher sensitivity and, provided you can read the raw data, a doubling of resolution. http://petapixel.com/2013/08/04/scratching-the-color-filter-array-layer-off-a-dslr-sensor-for-sharper-bw-photos/

Reply to this comment...


Two thoughts:

first, the bees sometimes bring in different sorts of pollen, its not 100% sorted although the bees "try to" sort.

second, I'd guess its much more accurate to look at the pollen samples through a microscope (you only need one single pollen) than to get a big enough and sort-clean enough pollen sample for spectroscopy. Otherwise a kind of PCR (ELISA-test) would be much more specific and accurate. Wouldn't it be better to start a pollen microscopic image or a pollen DNA database?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Forager bees work only 1 species of flower on any single flight, so when you collect individual pollen packets from returning forages they are monospecies. We will be collecting micrographs and DNA barcoding samples that have spectrographic data. This way we can positively assign species identifications to spectrographic fingerprints. BTW PCR and ELISA are both very different tests and ELISA wouldn't be of any help here.

Reply to this comment...


If you're at all interested in reflectance measurements, I posted a sketch for a reflectance setup here that might interest you: https://publiclab.org/notes/warren/04-21-2016/prototype-of-easy-to-build-rigid-spectrometer-housing#c14412

I probably won't have time to try it out but tell me if you do!

Reply to this comment...


Log in to comment