Photo-identification as a tool to study small-spotted catshark Scyliorhinus canicula

Image-Based Automatic Individual Identification of Fish without Obvious Patterns on the Body (Scale Pattern)

The precision fish farming concept has been widely investigated in research and is highly desirable in aquaculture as it creates opportunities for precisely controlling and monitoring fish cultivation processes and increasing fish welfare. The automatic identification of individual fish could be one of the keys to enabling individual fish treatment. In a previous study, we already demonstrated that the visible patterns on a fish’s body can be used for the non-invasive individual identification of fishes from the same species (with obvious skin patterns, such as salmonids) over long-term periods. The aim of this study was to verify the possibility of using fully-automatic non-invasive photo-identification of individual fish based on natural marks on the fish’s body without any obvious skin patterns. This approach is an alternative to stressful invasive tagging and marking techniques. Scale patterns on the body and operculum, as well as lateral line shapes, were used as discriminative features for the identification of individuals in a closed group of fish. We used two fish species: the European seabass Dicentrarchus labrax and the common carp Cyprinus carpio. The identification method was tested on four experimental data sets for each fish species: two separate short-term data sets (pattern variability test) and two long-term data sets (pattern stability test) for European seabass (300 individual fish) and common carp (32 individual fish). The accuracy of classification was 100% for both fish species in both the short-term and long-term experiments. According to these results, the methods used for automatic non-invasive image-based individual-fish identification can also be used for fish species without obvious skin patterns.

Computer vision based individual fish identification using skin dot pattern

Precision fish farming is an emerging concept in aquaculture research and industry, which combines new technologies and data processing methods to enable data-based decision making in fish farming. The concept is based on the automated monitoring of fish, infrastructure, and the environment ideally by contactless methods. The identification of individual fish of the same species within the cultivated group is critical for individualized treatment, biomass estimation and fish state determination. A few studies have shown that fish body patterns can be used for individual identification, but no system for the automation of this exists. We introduced a methodology for fully automatic Atlantic salmon (Salmo salar) individual identification according to the dot patterns on the skin. The method was tested for 328 individuals, with identification accuracy of 100%. We also studied the long-term stability of the patterns (aging) for individual identification over a period of 6 months. The identification accuracy was 100% for 30 fish (out of water images). The methodology can be adapted to any fish species with dot skin patterns. We proved that the methodology can be used as a non-invasive substitute for invasive fish tagging. The non-invasive fish identification opens new posiblities to maintain the fish individually and not as a fish school which is impossible with current invasive fish tagging.

Recognition software successfully aids the identification of individual small-spotted catsharks Scyliorhinus canicula during their first year of life

Eighteen captive small-spotted catsharks Scyliorhinus canicula were successfully identified from hatching to 1 year of age using the free computer recognition software, I³ S classic. The effect of increasing the time interval between recognition attempts on the accuracy of the software was investigated, revealing that recognition fiedelity decreases with increasing time intervals for younger (0 to 15 weeks), but not older (15 weeks onwards) sharks. Identification by I³ S was validated using genetic analyses of seven microsatellite markers, revealing a 100% success rate. Thus, this non-invasive recognition method can be used as an inexpensive and effective alternative to invasive tagging, improving animal welfare and complementing ex-situ conservation methods.