Extracting fish size using dual underwater cameras

The definition of a remote system in the monitoring of fin fish growth rate and shape change relies on the development of the appropriate optical ranging system and the automation of data collection. Among the available technologies, a dual camera optical ranging system is presented as the most suited for fish size remote estimation. Images are collected via a submergible low-priced dual camera module connected to a portable waterproof PC equipped with two frame grabbers. Two images are synchronically collected and, with the system parameters (focal length of both cameras, distance between the two cameras, vertical and horizontal tilt and relative rotation of the two cameras), are used to collect data of fish size and shape. A Neural Network is built to correct the error of measurement. A geometric algorithm is developed to filter fish images and elliptic Fourier analysis of automatically collected fish outline coordinates is proposed as a tool of shape analysis. For all other fish orientations, landmarks (homologous points) are collected on fish outlines and evident structures: landmark configurations relative to each fish are then rotated using the system parameters. The proposed system can be used in the monitoring of sea-based fish farming facilities, especially those which are permanently submerged, reducing mortality and stress due to fish sampling and limiting divers intervention. By its application in the study of wild population, the system can be useful in the characterization of fish communities and population dynamics, supporting traditional visual census techniques with a tool which allows for continuous, remote and automatic data collection.