A Review of Feature Extraction Technologies for Plankton Images

Abstract

Plankton image feature extraction technology is the fundamental basis of automatic plankton tracking, classification, and recognition. Various feature extraction technologies have been proposed and put into practice for different underwater imaging environments. This paper gives a comprehensive review of plankton image feature extraction technologies from the last 10 years. The main contributions of our work are as follows: (1) a comprehensive introduction to geometric, texture, and local feature extraction technologies for plankton images, and a comparative analysis of their adaptive capacity when these methods are applied to complex underwater environments, such as insufficient illumination, large amounts of suspended particles, high intensity noise, and other factors that prevent the normal transmission of light; (2) a guide to choosing the proper feature extraction method for specific plankton species and specific purposes, such as detection, tracking, classification, and recognition; (3) a detailed discussion of feature enhancement technologies focusing on image segmentation and polygonal approximation of plankton target contours. Finally, we review developing trends in underwater imaging equipment to predict future plankton feature extraction technologies, in hopes of helping scientific researchers in this field. We also introduce a set of real-time mobile terminal applications for plankton monitoring, identification, and classification based on a projection-based underwater sampling device called PlanktonScope, which is currently in use in our laboratory.