Deep learning has been applied to the problem of intelligent plankton classification, which could have important implications for understanding marine ecosystems, the food chain, and the environmental impact of oceanic microbes on climate.
Hussein Al-Barazanchi and Shawn Wang of California State University, in Fullerton and Abhishek Verma of New Jersey City University, Jersey City, USA, discuss the importance of plankton in the International Journal of Computational Vision and Robotics, and outline their intelligent plankton image classification.
Plankton is an umbrella term for any organism that lives in a large body of water, such as an ocean and cannot propel itself against the current. It is an extremely diverse group that encompasses bacteria, archaea, algae, protozoa and any drifting or floating animals that inhabit large water columns. Plankton is a source of food for fish and other marine animals. Moreover, the distribution of plankton underpins the persistence of marine ecosystems as well as having an impact on chemical concentrations of the oceans and the Earth’s atmosphere.
The team explains that because of the diversity of plankton in terms of their nature, size and shape, accurate classification is daunting and the mixed quality of images collected for different types of plankton and species makes this problem even more challenging.
The team’s new intelligent machine learning system based on convolutional neural networks (CNN) for plankton image classification does not depend on features engineering and can be efficiently extended to encompass new classes. Tests on standard images show the new approach to be more accurate than even state-of-the-art tools available today.
Al-Barazanchi, H., Verma, A. and Wang, S.X. (2018) ‘Intelligent plankton image classification with deep learning‘, Int. J. Computational Vision and Robotics, Vol. 8, No. 6, pp.561-571.