|
|
HYDROPTIC, 8 avenue du Commandant Taillefer, 31230 L' Isle en Dodon, Haute Garonne, France |
|
  Phone: 09 63 24 82 20 Fax: 05 61 89 37 88 EMAIL |
COMPUTER SOFTWARE
Introduction:
For image-processing/classification software to be useful to the world-wide zooplankton research community it must be capable of:
1) importing zooplankton sample images into the system followed by the isolation of individual organisms of interest,
2) providing a means of visually sorting individual vignettes to build a 'training set' consisting of groups of vignettes of organisms of the same type whose identities can be verified by a taxonomic expert,
3) using a variety of different classification methods and algorithms that can "learn" to classify unknown images,
4) allowing the development of standards for the presentation of metadata.
The open-source computer package
ZooPROCESS with Plankton Identifier is an example of such software which allows specific types of images to be processed and the degree to which each image classification is integrated into the system.
The
ZooSCAN and
UVP systems are delivered complete with a processing software suite on disk. All the necessary files can be extracted from the ZooSCAN zipped archive and this can also also be downloaded from the . Note that the versions of the software available from this archive may not be the latest available for each of the different software packages (e.g. Vuescan, EPSON) but these versions have been fully tested. For example, Vuescan versions 8.3.23 and 8.4.57 are the only Vuescan versions which are supported .
The
ZooPROCESS macros can be downloaded separately and extracted from the
ZooPROCESS zipped archive also available from the website.
Vuescan
Vuescan versions 8.3.23 and 8.4.57 are supported. Refer to the
ZooPROCESS manual to be found for detailed instructions.
ZooPROCESS
ZooPROCESS is a open-source and free software package written in
ImageJ macro language by Marc Picheral from LOV (CNRS). It was designed initially for the ZooSCAN system but can also be used by
UVP. It facilitates calibration of the system and acquires and archives the vignettes of zooplankton individuals and their metadata.
ZooPROCESS also allows image processing, analysis, and feature extraction. It isolates ROIs, permits manual classification, and supports the preparation of learning and testing data sets for classifiers.
Image classification is accomplished using the integrated
Plankton Identifier package which incorporates the Tanagra "drag-and-drop" data mining routines. This facilitates the building of 'training-set' databases against which species are identified.
Plankton Identifier was developed by Stephane Gasparin also from LOV (CNRS).
ZooPROCESS software is based on the ImageJ macro language.
ZooPROCESS also provides tools for quality control, and is linked to Plankton Identifier software. Plankton Identifier is free software for automatic and semi-automatic classification of plankton. For the
ZooSCAN application it is interfaced with the
ZooPROCESS software but it can also be used standalone. It has been developed in DELPHI (Borland) programming language because the source code can be compiled, giving applications with shorter execution time. This aspect is important when a large set of samples containing many objects has to be analyzed. Plankton Identifier works in conjunction with the free data mining software Tanagra, also developed in DELPHI.
Automatic and semi-automatic recognition of plankton
The system provides two kinds of data for scientific users: the abundance of different zooplankton groups and their size. The different zooplankton groups can be obtained by a fully automated classification or by a semi-automated one. The semi-automated method involves validation (correction) rather than the use of a detailed 'learning set' (prediction). The automated process substantially decreases the validation effort and is thus time-saving. The biologist has full control over the number of categories to validate and the objects that are assigned to the categories. Technicians, students and scientists can be rapidly trained in manual sorting using the ZooSCAN image atlas which can be found
The results obtained by the developers of ZooSCAN and
ZooPROCESS from a series of surveys are encouraging for the estimation of zooplankton size and biomass spectra. Many ecological traits (including metabolic rates, population abundance, growth rate and productivity, spatial habitat, trophic relationships) are correlated with body size. The ZooScan imaging system has provided an efficient means of reconstructing plankton size spectra from taxonomically well-characterized zooplankton samples. In addition, automated measurements of either linear or areal dimensions of digitized organisms has been related to their biomass. Plankton abundance size and biomass spectra are critical information for many biogeochemical models that include Plankton Functional Types or size based trophic interactions.
ZooSCAN 
