Rainbow trout in most of the proliferation and breeding sites of cold-water fishes has been propagated and inbred. One of the proliferation steps of this type of fishes is the separating fertile and living fish eggs from the infertile or dead ones and counting them for sale. In spite of various apparatuses and methods of proliferation, the recognition of fertile from dead fish eggs is essential. In this study, the ability of machine vision system coupled with soft computing methods such as Artificial Neural Networks (ANN) was examined to quality assessment of fish eggs. In this regard, the captured images were transferred to the LAB color domain, because this domain is less affected by the camera and lighting conditions then several color and textural features were extracted from the images of rainbow trout fish eggs. Finally, extracted features were introduced to ANN as an input layer. As a conclusion, results showed that with an optimum adjustment of ANN, the live and dead fish eggs were classified with 99% accuracy. The outcome of this investigation can be used in the fish egg quality assessment.
