Arribada Initiative is building an early animal detection and warning system to alert communities to the presence of dangerous animals. Our system relies on thermal array sensors capable of taking video of an animal thermal signature. We hope to train an algorithm to automatically recognize dangerous animals and send an alert to the community.
We are posting periodic progress updates on WILDLABS.net. Our first post can be found here: https://www.wildlabs.net/resources/case-studies/hwc-tech-challenge-thermopile-sensor-project
Our 2nd progress update should be posted on WILDLABS shortly.
We would love any feedback or suggestions from the community on our project!
30 August 2018 9:05pm
Hi Anne and the Arribada team. This is such an exciting project and so pleased to see the progress you are making.
A couple of quick ideas sprang to mind reading your latest update. Firstly, the Lepton images are actually quite impressive I reckon. I was very pleasently surprised as I'd never really seen performance in relation to animals.
One thing I've noticed in working with thermal equipment is that sometimes it can help to have a known heat source/reference temperature to kind of callibrate the sensor. I first discovered this using a hand-held Flir unit made for industrial applications in searching for koalas in trees in Australia. Although the temperature range of interest could be set on the unit (also very useful), I noticed it performed even better when I placed my finger a little bit in front of the lens. It seemed to have an effect of markedly reducing the background thermal noise and made it a bit easier to spot koalas. I've sinced wondered about something a little more sophisticated, perhaps set up at various ranges from imagers, able to produce a fairly constant, known reference temperature and whether this would even help that much. Thoughts?
Also, we have a Flir Vue Pro R, the "R" standing for radiometric, which essentially just means that every pixel records quite an accurate temperature, which I've honestly never even looked at before. Recently I was asked by a researcher to borrow the unit to see if she could count flying foxes. I wondered whether, along with the actual images, the radiometric data might be useful. For example, I can imagine an algorithm scanning images looking for "clumps" or groupings of target temp pixels. I believe the Lepton series has a model that offers radiometry, which you may already have, and I'd be interested in collaborating on this front (e.g. providing reference data from the Vue etc., or whatever you think could be useful).
Finally, I don't know whether it'd be of much use, but we have an old Flir 120 x 80 pixel surveillance camera just sitting here that you are welcome to borrow. If you could cover shipping you are certainly welcome to use it for the term of the project. Not doing anything of value here at the moment.
Anyhow, my best for this great project and really looking forward to more updates.
Rob
Rob Appleby
Wild Spy