Application of laser speckles and deep learning in discrimination of supermicroplastics and zooplanktons

Plastics entering the ocean are further degraded by waves and UV rays, resulting in fragments. Plastic fragments whose size are equal to or less than 5 mm and 350 µm are called microplastics (MPs) and supermicroplastics (SMPs) respectively. Aquatic lives with accumulated MPs are reported to develop adverse effects. It is necessary to survey the amount and distribution, especially SMPs in the sea, to estimate the impact of MPs and SMPs on organisms. One of the current methods for investigating MPs is a towing net that has a size larger than about 350 µm mesh which makes it challenging to collect SMPs and thus lack information of SMPs. Therefore, we propose a new method of surveying based on laser speckles. The method uses two submersible spheres with a 635 nm laser and a CMOS camera mounted within each of the spheres and analyzing the recorded speckle images from the light scattered within the volume between the spheres. For the simulation studies, different sized polystyrene spheres (2 µm, 20 µm and 200 µm) and zooplanktons at certain definite concentrations were introduced into the volume to produce speckles. Speckles were analyzed using an algorithm of calculating difference frames of recorded video and deep learning algorithms to distinguish particles from planktons and particles of different sizes. Deep learning was found to be capable of distinguishing different particle sizes from the speckle patterns. In future, we will explore the potential of deep learning tools in the detection capability as a function of particle concentration in the presence of zooplanktons.