Date of this Version
We have developed a prototype web-accessible content-based image retrieval (CBIR) system that allows internet/web-based sharing of biological collections that contain large numbers of images of archived specimens. This system will enable both researchers and educators to access verified, high quality data on biological collections that are available in any museum with digitized holdings. The CBIR system that we are testing can play an important role in understanding global biodiversity because no knowledge of the specific names of specimens need be known before useful information can be extracted from such databases. Our CBIR framework allows users to search image collections using query by content, query by example, and query by sketch. Additionally, we have developed a web based administrative interface to maintain and monitor any image collection. A key idea of our project is to develop a framework that assists educators and researchers to identify biological specimens and to study biodiversity with greater precision and speed. This is a web-based system that is low cost, extensible, flexible, and easily configurable. Our CBIR prototype was implemented and tested using specimens of the family Opecoelidae (trematodes of marine fishes) that are housed in the Harold W. Manter Laboratory of Parasitology Collection at the University of Nebraska. In the future, we plan to field-test the prototype and assess the ability to provide image retrieval from a variety of biological images and with a multitude of image features.