Sub-second analysis of fish behavior using a novel computer-vision system

Abstract This work presents an advanced version of a previous computer vision system that is appropriate for analyzing more complex fish behavioral traits. The system is capable of long-term recording of fish escape and bite behavior in tanks with excellent sampling accuracy and a minimum number of frames lost. In addition, the system is able to simultaneously record from nine different tanks with nine respective cameras, thereby allowing for specific experimental designs for statistical purposes. The evaluation of the system's operation and capabilities is achieved under specific biological activities in laboratory experimental conditions, with the activities duration similar to the system time characteristics. A sub-second analysis resulted in a detailed description of the escape and bite patterns of sea bream and bass in discrete steps, according to the video sequences. In general, the system was found to be able to assist in performing the behavioral studies of farmed fish. The final cost-effective system is characterized by long recording periods of high sampling accuracy, multiple digital camera acquisitions with high image quality, and state-of-the-art consistency.