Detritus identification in FlowCAM using a simple binary classifier
Garcia-Oliva, O.
Show abstract
Phytoplankton and detritus particles may be co-captured in Flow-CAM systems, leading to misrepresentation of phytoplankton abundance. In this study, I use logistic regression as a binary classifier to identify detritus particles in FlowCAM data from the Southern North Sea. Standard particle features from the manufacturers software were used as inputs, with surface texture (intensity variance) and compactness (derived from particle perimeter and area) proving most relevant to detritus classification. This classifier achieved 81% accuracy using a training dataset of approximately 7300 observations, reducing the workload compared to other classification methods. The reconstructed particle size spectra closely matched the observed spectra for detritus and phytoplankton. Binary classifiers like this offer a fast, effective alternative for detritus screening, aiding the pre-processing and re-analysis of FlowCAM datasets.
Matching journals
The top 3 journals account for 50% of the predicted probability mass.