Automatic Recognition Method of Zooplankton Image in Dark Field

  • Ying Tian
Keywords: Dark Field, Zooplankton Image, Automatic Image Recognition, Feature Extraction

Abstract

Marine zooplankton biomass, population structure, community diversity and migration processes play
an important role in marine ecosystems, marine biogeochemical cycles, marine environments and global climate
change research. The massive image data generated by the in situ plankton optical imaging observation system
is a challenging problem in marine plankton survey. Based on the existing identification methods, this paper
carries out image preprocessing of the dark field zooplankton population. Research on processing methods such
as feature extraction and classification recognition laid the foundation for constructing an automatic recognition
system for zooplankton in dark field. In the aspect of image preprocessing, the histogram equalization is firstly
used to reduce the image quality caused by uneven illumination, and the adaptive wavelet method is used to
suppress the underwater zooplankton image noise while retaining the original zooplankton details. Secondly, the
area detection is adopted. The algorithm extracts the region of interest from the whole image, and proposes a
watershed segmentation method based on morphological reconstruction, and designs a two-level classifier
strategy to further improve the recognition accuracy. The results show that the recognition accuracy of the twolevel
classifier is better than that of a single classifier.

Published
2019-11-01