Development of an automated computer tool for bee identification from wing venation

Monday, November 16, 2015
Exhibit Hall BC (Convention Center)
Emad Sadeghi , Dept. Electrical and Computer Engineering, University of Wisconsin - Madison, Madison, WI
William Sethares , Dept. Electrical and Computer Engineering, University of Wisconsin - Madison, Madison, WI
Claudio Gratton , Department of Entomology, University of Wisconsin-Madison, Madison, WI
The ability to accurately catalog and enumerate species is key to monitoring and conservation

planning. However, species identification for difficult groups such as insects has largely been

the purview of a small number of specialized taxonomists. We present early results

of a computer-assisted tool developed for species-level identification of bees. The computer

software, which is easy to use, utilizes advanced image processing algorithms for the rapid and

accurate identification of bees from wing venation. To attain accurate results, the visual

characteristics of various bees groups are to be extracted. This is done by training the

embedded computer algorithms via the labeled images of a known group of bees. To evaluate

the accuracy of the developed algorithms, extensive tests have been done on unlabeled native

bee species of Wisconsin. The results show a species-level identification of bees with up to 95%

accuracy.

Previous Poster | Next Poster >>