Fig 1.
An example of a raw image collected by the ZOOplankton VISualization system in Chesapeake Bay, showing the large amount of particulates along with an arrow worm.
Fig 2.
An example of a raw image collected by the ZOOplankton VISualization (ZOOVIS) system in Chesapeake Bay, and results from various thresholding approaches: A ZOOVIS image showing a hydromedusa, copepods and other particulates. B Binary image from the global threshold approach showing the amount of segmented objects, C Binary image from Maximally Stable Extremal Regions (MSER) approach showing hydromedusa and other large particulates, D Binary image from the customized adaptive thresholding approach showing copepods and other small particulates, and E Final combined binary image with red boxes showing the segmented objects.
Fig 3.
Examples of model species and construction of feature descriptor.
A—C: Examples of segmented arrow-like, copepod-like, and gelatinous zooplankton, D—E: standardized (250 by 250 pixels) denoised cropped images, and F—H: Histogram of Oriented Gradients feature descriptors for three categories.
Fig 4.
A schematic flow chart showing the proposed image processing procedure.
Fig 5.
Example of copepods extracted with different sizes and orientations using the computer from the semi-automated procedure.
Note the cropped images were denoised and then classified by the classifier.
Fig 6.
Example of ctenophores and hydromedusae with different shapes and sizes extracted using the semi-automated procedure.
Note the cropped images were denoised and then classified by the classifier.
Fig 7.
An example of image comparing visual counts and results from computer-based procedure.
A) A raw image contains copepods showing copepods in focus indicated by red boxes and out of focus indicated by yellow boxes. B) The same image processed by the computer-based procedure with red boxes indicating segmented objects and resulting copepods were numbered.