Upload

Loading icon Loading...

This video is unavailable.

idBee: Automated Bee Identification from Wing Venation

Sign in to YouTube

Sign in with your Google Account (YouTube, Google+, Gmail, Orkut, Picasa, or Chrome) to like EngineeringUW's video.

Sign in to YouTube

Sign in with your Google Account (YouTube, Google+, Gmail, Orkut, Picasa, or Chrome) to dislike EngineeringUW's video.

Sign in to YouTube

Sign in with your Google Account (YouTube, Google+, Gmail, Orkut, Picasa, or Chrome) to add EngineeringUW's video to your playlist.

Published on May 30, 2012

http://idBee.ece.wisc.edu

About 35% of the world's crops and 80% of the world's flowers are pollinated by bees. Recently, honeybees have begun dying off at an alarming rate, and no one is quite sure why; researchers call it Colony Collapse Disorder. Learning more about local bees is crucial because there is a chance that these local bees may be able to take over some of the pollination duties of the honeybees. This involves tracking wild bee populations, abilities, and habits. There are many species of bees, more than 500 in Wisconsin alone, and it is not easy to tell (just by looking) which species an individual belongs to.

Rather than capturing, killing, and sending the bees off to be catalogued, imagine an iPhone app that lets graduate students and researchers identify bees in the field. One could simply take a photo of a bee and instantly record its species and location. Research could be conducted much faster, and identified bees could be released back to nature.

This is not a reality yet, but we're taking the first steps. We call it "Google for Bees."

Here's how we envision it working. Researchers go to our website, (or ultimately an iPhone-style app), and provide a photo of a bee wing to match against a species the system has learned. The system applies image processing filters to determine the outline of the wing and veins and then determines the species by matching the description of the bee wing with those of species that it has already learned.

The foundational elements of the system have already been developed. Our test system can identify about 20 different bee species with over 90% accuracy. But, we need help to improve the system.

Currently, photos need to be taken with sophisticated lab cameras, with proper lighting, and with proper placement of the wing in the photo. To truly be valuable to bee researchers, the system needs to be able to identify many more species from photos taken with lower-quality cameras under less precise lighting and positioning conditions. The system could be further enhanced with the additional capability of identifying the gender of the bees.

We can't make the "Google for Bees" without collaborating with researchers outside of Engineering. We need many more bee-wing photos to improve the performance of our image processing and enhance our machine learning algorithms, and we need to know the species depicted in each photo (during the training phase). For example, we estimate that we need 20 to 30 photos of different bees within each new species in order to reliably detect that species in the future -- with 500 local species, this is a lot of bees and photos. With 20,000 bee species world-wide, it is an immense task. Finally, we need researchers to test the system to make sure that it continues to work reliably.

  • Category

  • License

    Standard YouTube License

Loading icon Loading...

Loading icon Loading...

Loading icon Loading...

The interactive transcript could not be loaded.

Loading icon Loading...

Loading icon Loading...

Ratings have been disabled for this video.
Rating is available when the video has been rented.
This feature is not available right now. Please try again later.

Loading icon Loading...

Loading...
Working...
Sign in to add this to Watch Later

Add to