In this study a visual machine technology-based intelligent system was developed and evaluated for separation and recognizing the alive and dead eggs of rainbow trout fish. The features derived from imagery processing of alive and dead eggs were used as the decision-making variables in the classifier. Multi-layer Perceptron neural network -MLP- and Support Vector Machine -SVM- models were used as the classifiers. With paired t-test, 10 effective features were selected from 15 features for classification. The k-fold cross validation method was used for better evaluation the classifiers. By changing the size of the training data set from 80% to 20%, the classifier ability and stability were evaluated. The results showed that in the training phase, all the mean values of the statistical indices for MLP and SVM classifications were complete for all categories -100% of the classification was predicted correctly-. Also, in the test phase, the performance indicators of both classifiers were very satisfactory -the average accuracy was 99.45%-. Therefore, it is possible to use both classifiers with certainty for separation the rainbow trout fish eggs.

Keywords