Fish length measurements using artificial neural networks

Abstract

An existing stereoscopic technique employing neural networks has been used to measure the length of fishes. Prior to the actual measurements, certain parameters that might affect the accuracy of the measurement were investigated. The influence of the index of refraction of water (depending on salinity) and the orientation of the object relative to the cameras on the accuracy of the measurement was examined.

Results showed that the salinity of water and the orientation of the object with respect to the. cameras have negligible effect on the measurements. With a total error of less than 2 mm, the method presented in this paper is far better than conventional techniques.