zooplankton: Zooplankton image analysis

In SciViews/data.io: Read and Write Data in Different Formats

Zooplankton image analysis

Description

Various features measured by image analysis with the package zooimage and ImageJ on samples of zooplankton originating from Tulear, Madagascar. The taxonomic classification is also provided in the class variable.

Usage

zooplankton

Format

A data frame with 19 variables:

ecd

The "equivalent circular diameter", the diameter of a circle with the same area as the particle (in mm).

area

The area of the particle on the image (in mm^2).

perimeter

The perimeter of the particle (in mm).

feret

The Feret diameter, that is, the largest measured diameter of the particle on the image (mm).

major

The major axis of the ellipsoid matching the particle (mm).

minor

The minor axis of the same ellipsoid (mm).

mean

The mean value of the gray levels calibrated in optical density (OD), thus, unitless.

mode

The most frequent gray level in that particle in OD.

min

The most transparent part in OD.

max

The most opaque part in OD.

std_dev

The standard deviation of the OD distribution inside the particle.

range

Transparency range as max - min.

size

The mean diameter of the particle, as the average of minor and major (mm).

aspect

Aspect ratio of the particle as minor/major.

elongation

The area divided by the area of a circle of the same perimeter of the particle.

compactness

sqrt((4/pi) * area) / major.

transparency

1 - (ecd - size).

circularity

4pi(area / perimeter^2).

density

Density integrate by the surface covered by each gray level, i.e. O.D., inside the particle.

class

The classification of this particle. 17 classes are made. Note that Copepods are Calanoid + Cyclopoid + Harpactivoid + Poecilostomatoid and they represent the most abundant zooplankton at sea.

This is a typical training set used to train a plankton classifier with machine learning algorithms. Organisms originate from various samples (different seasons, depth, etc. to take the variability into account). However, the abundance of the different classes do not match abundance found in each sample, i.e., rare classes are over-represented in this training set. Only zooplankton classes are present in the dataset. Full data also contains classes for phytoplankton, marine snow, etc. Take care that several variables are correlated!

Source

Grosjean, Ph & K. Denis (2004). Supervised classification of images, applied to plankton samples using R and ZooImage. Chap.12 of Data Mining Applications with R. Zhao, Y. & Y. Cen (eds). Elsevier. Pp 331-365. https://doi.org/10.1016/C2012-0-00333-X.

Examples

table(zooplankton$class)
library(ggplot2)
ggplot(zooplankton, aes(circularity, transparency, color = class)) +
  geom_point()

SciViews/data.io documentation built on May 5, 2024, 1:39 p.m.