Bilinear Convolutional Neural Network with Random Projection Paradigm based Classification of Micro Plastics in Marine Environments

In recent years, the immense growth of micro plastic pollution in the marine environment increased the attentions globally. Micro plastics particles referred as the chemical substances and polymers based on different types of toxic substances the pollutes the marine water when it decomposed into monomers. Thus, in existing researches, machine learning algorithms utilized for classification, however, these methods have drawbacks of misclassification due to variation in appearance of the plastic compounds which are in micro size. To overcome this drawback, a Bilinear Convolutional Neural Network with Random Projection Paradigm (BCNN-RPP) method is proposed for precise classification of micro plastics in marine environment. The proposed BCNN-RPP model obtained the texture variations of the plastic compounds efficiently which are crucial is differentiating the various kinds of micro plastics. The RPP technique improve the ability of the proposed BCNN model to classify the complex plastic compounds accurately. The experimental results of the proposed BCNN-RPP achieved high accuracy than the existing classification approaches like Mask Region-Convolutional Neural Network (Mask-R-CNN).