IMAGE ANALYSIS AND RESEARCH
BY SAMPLE AND CONTENT
ON THE INTERNET
Introduction
On one hand the World Wide Web makes up a considerable information system including texts, sound, pictures and videos, which are available to every one, every computer which is connected with the internet.
On the other hand to give you an idea of the volume of data which are manipulated on a day to day basis, an earth observation satellite for civil use can produce up to 50000 images per year . An image uses from 30 to 100 mega octets. If you calculate ten years for a series of four satellites, a million billion octets are easily reached. This necessitates large storage centres and excellent data base management for indexation and retrieval.
In order to accede effectively
to the information present in a storage centre, a query for information:
Firstly, an automatic indexing showing the content of information
of image and video must be organized.
Secondly, a information query must be broken down into several
resolution levels to have access to data with a high degree of
flexibility:
• Access through the image’s address in the storage unit’s static data organization)
• Access by example of the recorded data using a multi Scritical selection applied to precalculated catalogue units (pseudo-static data organization)
• Relational access expressed in interrogation language of the user post’s data base (dynamic data organization) that is the access by the image content analysis.
I - General presentation
1-1 The Société de Bio-Informatique et de Bio-Technologie
The Société de Bio-Informatique et de Bio-Technologie is a research and development company whose aim it is :
- To promote technology transfer in the field of computing( picture treatment etc.),bio computing, bio-technologies and to promote technology transfers in the field of Internet.
- To apply them in the following fields: medicine, biology, chemistry, agriculture, aero spatial
1-2 The project
The project is the continuation of a previous work realized in the frame of the PACHA project of ESPRIT program (EC) on the automatic images analysis through the example and the content. From PASHA studies the project objective is the development and diffusion of a content based on images retrieval system on the Internet and for industrial medical applications.
This system will permit after automatic image indexing :
1-2.1 Retrieval image by sample
The
operator presents for example a picture of Lunar crater with the
means of a scanner and searches in the image basis( same dimension,
same age, etc…), all crater images are similar. The retrieval
gives thanks to the found images the known characteristics.
1-2.2 Retrieval image content analysis
The retrieval image content analysis by means of the query system is using artificial intelligence tools. The operator describes the object he wants to search in the image basis: Lunar crater:
with diameter > 200 km
with depth > 100m
with age > 1 000 000 years
with central piton
and the system will retrieve all craters having approximately
these characteristics (planned expositions in museums, technology
parks , astronomic observatory, etc….)
 |
 |
The
system permits also to retrieve approximately the same configuration
of objects:
Lunar
crater near to a valley depth > 300m
near
to mountains height > 1000m
An expert system will analyse the query and the pilot will analyses
the tools permitting the retrieval.
1-2-3 The team and the partners
This project has been studied in the in the Company by
- Claude GAUDEAU (Statistics and Mathematics Diploma, Doctor in Human Biology , professor at the Conservatoire des Arts et Métiers of Tours)Director of the Company.
- Phillippe BOBOLA( doctor in Physics and Chemistry, Scientific Assistant)
- Eric ROBERT( Mathematician Scientific AssistantWith the following partners:
- ADERSA, Mr Olivier GUYE- Palaiseau- FRANCE
- The « Ecole Supériuere d’Electricité, Pr TIMSIT- Orsay-France
- The Institut Universitaire de Technologie, Departement Génie Eléctric et Informatique Industrielle, ROUSELLE-Avenue Monge- Parc de Grammont 37200 Tours.