Automated counting of phytoplankton by pattern recognition: a comparison with a manual counting method

Abstract

Computer-based image analysis and pattern recognition methods were used to construct a system able automatically to identify, count and measure selected groups of phytoplankton. An image analysis algorithm was employed to isolate and measure objects from digitized images of a phytoplankton sample. The measurements obtained were used to identify selected groups of phytoplankton by a combination of artificial neural networks and simple rule-based procedures. The system was trained and tested using samples of lake water covering an annual growth, cycle from Lough Neagh in Northern Ireland. Total volume estimates were obtained for the four major phytoplankton species, using both the automated sjstem and a manual counting method. Estimates of total cell volume obtained from the automated system were within 10% of those derived by manual analysis of the same cells. The automated system produced total cell volume estimates dose to those obtained from manual analysis of different aliquots of the same water sample. Variation between successive counts of the same water sample was higher with the automated system than with the manual counting method. Limitations and possible improvements to the technology are discussed.