Automatic In Situ Identification of Plankton

1
Automatic In Situ Identification of Plankton
M. Blaschko, G. Holness, M. Mattar, D. Lisin, P.
Utgoff, A. Hanson, H. Schultz, E. Riseman, M.
Seraki, W. Balch, B. Tupper

• Motivation
• Phytoplankton is the basis of the food chain for
  marine life
• Integral component of global carbon cycle
• Studying abundance of different species
  important for understanding of global and
  local ecology
• Manual identification is a daunting task, so
  automated solution is needed

• Problem
• Identify taxa of phytoplankton from images
  taken in situ by FlowCAM

System Overview
Classification

• Classification Methods
• K-Nearest Neighbors
• Decision Trees
• Naïve Bayes
• Ridge Regression
• Support Vector Machines

Ensembles

• Combined estimates can lead to increased
  accuracy
• Improvements possible if individual
  classifiers are independent
• Methods used Boosting, Bagging, and
  Multi-Classifier

• Instance x1 ltx11,,x1dgt
• Class label Yi ? c1,,cK class labels
• Labeled instance (xi, yi)
• Training set T (x1,y1),,(xN,yN)
• Partition feature space into regions
• Each region contains instances in a class
• Classifier Induction Estimate functionmapping
  instances to class labels
• Sometimes estimates commit errors

Single classifier Results

• Conclusion
• Combinations of shape and texture performed
  best
• Best results with Support Vector Machines Best
  accuracy was 73, comparable to consistency
  rate of human experts

• Experiments
• 980 expert labeled FlowCAM image
• pool of 780 total features
• 10-Fold Cross Validation

• Future Work
• Automated Feature Selection
• Improved ensemble performance gains by
  inducing classifier independence
• Experiments with Local image Features