Adaptation of algorithms for automatic recognition and classification of
plankton
Position at : Laboratoire dOcéanographie de Villefranche (LOV-UMR 7093),
B.P., 06234Villefranche sur mer, FRANCE.
In the framework of the studies concerning the impact of the climatic
change on the aquatic ecosystems, the long term series analysis in the
different geographic provinces is essential. Species or species assemblages
of plankton are good indicators of regime shifts and modifications in their
community structure or in their population dynamics are used both in
studies concerning the response of the ecosystem to environmental forcings,
and in predictive modelling of the processes related to these changes. The
Laboratoire dOceanographie de Villefranche (LOV) is one of the European
forerunners of the application of zooplankton imaging techniques at sea and
in the laboratory. Several instruments based on these methods were built
and are in use. The technology of the Underwater Video Profiler and of the
Zooscan was transferred towards the SMEs. Nevertheless, the treatment of
the data and their analyses remain a scientific activity that is carried
out in LOV.
There is a need for a classification system capable of handling images from
a variety of in situ and laboratory imaging systems; providing users with a
broad selection of classification algorithms; and classifying images with a
high degree of accuracy that is at least comparable to the performance of a
human expert.
The importance of such system is :
1) in the reduction of the delays of analyses in the treatment of long time
series,
2) the construction of classifiers allowing rapid assessment of the
temporal and spatial evolution of the functional groups of plankton, and
3) the creation of networks for experts interaction.
In order to fully exploit the imaging instruments, and to automate the
treatments, we must reinforce our competences in the treatments of images
sensu stricto and develop feature extraction tools and algorithms for
recognition and classification of the objects. It is imperative to develop
algorithms allowing the identification and analysis of complex natural
objects in a noisy environment.
The objective of this project is the development of approaches allowing the
implementation of algorithms for image treatment and analysis of
descriptive parameters which will be introduced into neural network and
numerical analyses for the final classification. At first, these methods
will be applied on the long time series of the Mediterranean Sea.
Candidates profil:
This proposition concerns specialists in the image treatment and analysis
software development. The aim is the adaptation of intelligent systems
for the automated recognition and classification of aquatic organisms.
Skills of the candidate are wished in following domains:
application of neural network and classifiers construction;
development of classification algorithms;
development of algorithms for extraction of the descriptive attributes of
objects;
development of applications to assess the quality of objects and their
morphological structure;
detection of objects in a noisy environment;
image treatment (correction, filtration techniques, arithmetical, logical
and geometric treatments);
image analysis (histogram, particles analysis);
image compression and storage
Contact :
Dr. Gabriel Gorsky (
[hidden email])
M. Marc Picheral (
[hidden email])
Information :
www.zooscan.com
Read the condition for a successful candidature :
https://www2.cnrs.fr/DRH/post-docs07/Download the candidature form :
https://www2.cnrs.fr/DRH/post-docs07/medias/tpl/pdf/commun/dos-candidat_post-docs07_fr.pdf
LAB website :
http://www.obs-vlfr.fr/LOV/
Locked
