Automating insect identification: exploring the limitations of a prototypesystem

Automated identification systems based on computer image analysis technolog y provide an attractive, but as yet unexploited potential solution to the g rowing burden of routine species identifications presently faced by a dwind ling community of expert taxonomists. DAISY (the Digital Automated Identifi cation system) is a prototype novel automated identification system, develo ped to explore this possibility. In its pilot phase, the DAISY algorithms w ere developed to discriminate five species of parasitic wasp, based on diff erences in their wing structure. Here, again using wing form, the ability o f DAISY to discriminate amongst an order of magnitude more species - 49 spe cies of closely related biting midges is examined. In so doing an attempt i s made to establish a set of basic 'benchmark' tests of the efficacy, and w eaknesses, of such an automated identification system.