In Environmental monitoring and assessment
The aquatic macroinvertebrate community reflects the ecological status of a river. Typically, some extraction methods have been implemented, but the capture and preservation of organisms are necessary. The techniques of digital image processing applied to ecology have become innovative tools for the characterization of aquatic macroinvertebrates. This research implements a methodology for the processing and classification of four aquatic macroinvertebrates genera Thraulodes, Traverella (Ephemeroptera), Anacroneuria (Plecoptera), and Smicridea (Trichoptera) present in three rivers in Antioquia (Colombia), which includes two phases. The first of these was the collection and capture of organisms to obtain a database of the most abundant genera, at laboratory scale. The second was the use of simulations that allow the classification of data through a process of selection and extraction of characteristics using the bag of visual words technique. Of all the classifiers tested, Gaussian vector support machines obtained a percentage of success in the recognition up method of four organisms to the genus level of 97.1 %. The training and computational processing for classification enabled the standardization of an appropriate methodology that will serve as a starting point for aquatic biomonitoring and inventory in Colombia and internationally.
Serna López Juan Pablo, Fernández Mc Cann David Stephen, Vélez Macías Fabio de Jesús, Aguirre Ramírez Néstor Jaime
Aquatic macroinvertebrates, Digital image processing, Machine learning, Vector support machines, Water quality monitoring