Automatic individual non-invasive photo-identification of fish (Sumatra barb Puntigrus tetrazona) using visible patterns on a body

Abstract

Non-invasive fish identification of individuals can provide new possibilities for the monitoring of fish cultivation, improve and make fish production technologies less demanding for farmers, and increase fish welfare. The aim of this research is to confirm the idea of automatic non-invasive image-based fish identification of individuals using visible features on a fish body and prove the pattern stability during the fish cultivation period. Visible patterns, such as black stripes along the body of a Sumatra barb (Puntigrus tetrazona), were used for machine identification of individual fish. Two experiments were completed: a short-term experiment (43 fish) to show the uniqueness of the stripe patterns for identification, and a long-term experiment (25 fish) to test the stability of patterns during the cultivation period. The overall accuracy of classification was 100% for data collection in one day and 88% between two data collection times. This study shows that visible patterns and image processing methods can be used to automatically identify individual fish of the same species. This is not just limited to Sumatra barb—the concept should work for any fish with unique visible skin patterns, for example, for commercial fish species like Atlantic salmon (Salmo salar) and European perch (Perca fluviatilis).