Publication details
Evaluation of the use of Artificial Neural Networks in Taxonomy: towards automating insect identification.
| Authors | |
|---|---|
| Year of publication | 2006 |
| Type | Article in Proceedings |
| Conference | Proceedings of the 10th International Conference on Chemometrics in Analytical Chemistry: CHEMOMETRICS IN THE TROPICS: Nature, Medicine and Industry CAC-2006 |
| MU Faculty or unit | |
| Citation | |
| Field | Zoology |
| Keywords | ANN; Insects identification; tachinids; thrips; jumping plant-lice |
| Description | In contrast to wide applications in chemistry, the use of ANN in taxonomy is rather rare, e.g. in chemotaxonomic identification of limpets or bioacoustics identification of Orthoptera, even if visionary study was published already in 1997 by Weeks 3. Perhaps the first real entomological application was used in the family Psychodidae (Diptera). Recently, we are building-up ANN methodology for insect identification. Appropriate key morphological characters (input) for species and utilized specimens correctly classified are creating database. With ANN we are finding model between input and species (output). In contradiction to manual identification, all characters are simultaneously taken into account over the complete database. ANN approach was developed, tested and applied in various species from three different insect orders: Diptera (Tachinidae), Thysanoptera (Thripidae) and Hemiptera (Psylloidea). Concluding, methodology developed is quite general and can be used for all entomological objects where sufficient number of characters is available and create the appropriate database After ANN “learning” the identification is fast and reliable. The approach is non-destructive unlike e.g. molecular analyses. |
| Related projects: |