CPICS: Continuous Particle Imaging and Classification System

The CPICS instrument has produced unprecedented results in the in situ aquatic microscopy of seawater, freshwater and laboratory samples. Using darkfield illumination, the CPICS captures high-resolution color images, showing features as small as 1µm. Research has shown that color information is key to high-accuracy classification while also providing important physiological information such as pigmentation due to grazing on phytoplankton. Because of its open-flow approach to water sampling, the delicate structures of plankton and particles remain completely intact as do predator-prey interactions.

Aquisition and Classification Software

The CPICS generates fullframe images at 6-10 Hz/s. Target ROIs (regions of interest) are extracted and stored for further analysis and archival purposes.

Features are extracted from the recorded ROIs and go through an automated classification process to generate taxanomic IDs for each ROI. The automated process is trained by manual identification of ROIs to form a training set.

roiSorter Software

For manually sorting ROIs into trainingsets for the automated classification.

"Crab Zoea" From Automated Classification


  • Tiwari, S. and SM Gallager. Machine learning and multiscale methods in the identification of bivalve larvae. Proc. 9th International Conference on Computer Vision, Beijing, China.
  • Tiwari, S and SM Gallager. 2003. Optimizing multiscale invariants for the identification of bivalve larvae. Proceedings of the 2003 IEEE International Conference on Image Processing, Barcelona, Spain, September 14--17, 2003
  • Thompson, C, M Hare, S Gallager. 2012. Semi-automated image analysis for the  identification of bivalve larvae from a Cape Cod Estuary. L&O Methods 10:538-554
  • Gallager, SM and S Tiwari, 2008 Optical method and system for rapid identification of multiple refractive index materials using multiscale texture and color invariants. Patent #60528965


For questions about CPICS please contact Scott Gallager

Regions of Interest (ROIs)

Poster (pdf) of regions of interest generated by a CPICS.

3D model of CPICS

Example Training Set

A subset of ROIs are manually sorted into training sets to feed into automated classification of the entire dataset.

Woods Hole Oceanographic Institution
©2013. Ocean Cubes, Woods Hole Oceanographic Institution. All Rights reserved. 266 Woods Hole Road, Woods Hole, MA | oceancubes@whoi.edu | 508 289-2783