Deep Classification of Microplastics Through Image Fusion Techniques

Microplastics from fiber shredding are recognized by the scientific community as one of the main sources of microplastic water pollution. Therefore, there is a compelling need for techniques capable of accurately identifying shredded microplastics in water. The recently released Holography Micro-Plastic Dataset, obtained through the use of digital holography microscope techniques, offers the opportunity to test the capability of deep neural networks to distinguish between microplastics and other debris on a standard benchmark. The promising results obtained from the initial batch of experiments can be further improved by employing a combined approach involving different image mapping techniques and leveraging recent state-of-the-art deep learning models. Within this framework, we analyze various image fusion schemes to merge the paired dataset images (amplitude and phase grayscale images) into a single three-channel picture. We demonstrate that our proposed approach yields increased accuracy compared to both single-image data processing and other fusion techniques. Finally, the performance of our method is further enhanced by utilizing the DenseNet model as the backbone for deep learning-based microplastic classification.