Generating Color Vignettes and Metadata for Insect Auto-Classification

Jordan Gimmen, Shermaine Garcia

Abstract


Classification of insects in biodiversity studies, where the variety of life forms within a particular ecosystem is measured, is often tedious and time consuming requiring the efforts of a trained taxonomist and many countless hours of work behind a microscope. Auto-classification has been attempted with success with zooplankton using ZooImage. Since plankton is transparent, ZooImage does not use color features when classifying. It produces metadata and vignettes (small images created from a larger source image) from a grayscale picture. Modifications to the program were also made to produce colored vignettes instead of grayscale vignettes so that color distinctions are displayed more evidently. After creating the modification, insect samples from a biodiversity study are processed by both the unmodified and modified programs. The two processes are compared to see which is more accurate. It is expected that the color features will be helpful in increasing the accuracy and reducing the number of false positives with the auto-classification. With the large number of insects that can be found in any single ecosystem, any effort to automate part of the process can incur substantial time savings.


Keywords


Insect Classification; Color Vignette; ZooImage; Computational Entomology

Full Text: PDF

Refbacks

  • There are currently no refbacks.


The Proceedings is produced as a service of UNC Asheville.