Application Research of Machine Learning in Automatic Identification of Fish Parasitic Viscera Species

In this paper, image processing and machine learning methods were applied to study how to classify and automatically identify the images of fish parasitic Myxosporidian cells. The shape feature and texture feature of the pre-processed image were extracted to generate the feature vector, followed by forming of the training set and the test set. The classification model was obtained by hyperparameter experiments of the three classifiers, and multiple indicators were used to evaluate the model. The accuracy of the Multi-Layer Perception (MLP) classifier reached 98.89% in the five types of fish parasitic Myxosporidian cells.