Multispectral Imaging for Automated Fish Quality Grading

Dhananjaya Jayasundara, Lakshitha Ramanayake, Neranjan Senarath, Sanjaya Herath, Roshan Godaliyadda, Parakrama Ekanayake, Vijitha R. Herath, Sujeewa Ariyawansha

Abstract: Fish grading is a vital process in the fisheries industry. In this paper, an algorithm is proposed utilizing multispectral imaging to automate fish grading. The images are obtained using an in-house developed Multispectral Imaging System. A Convolutional Neural Network (CNN) for image classification is utilized. From the CNN method, 93% accuracy was achieved. In addition to that, machine learning algorithms including Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) and Support Vector Machine (SVM) were performed on the preprocessed dataset for comparison purpose.