CUMULI: Computational understanding of multiple images

• Unité de recherche: PCRD EU
• Numéro de projet: 96.0095
• Titre du projet anglais: CUMULI: Computational understanding of multiple images

Mots-clé (Anglais)

Camera calibration; 3D vision; photogrammetry; CCD; Tracking

Autre Numéro de projet (Anglais)

EU project number: EP 21.914

Programme de recherche (Anglais)

EU-programme: 4. Frame Research Programme - 1.3 Telematic systems

Description succincte (Anglais)

See abstract

Partenaires et organisations internationales (Anglais)

Coordinator: INRIA Rhône-Alpes, Grenoble (F)

Résumé des résultats (Abstract) (Anglais)

Imetric, being a manufacturer of high-accuracy 3D Image Metrology Systems, works in a special framework. Typically cameras have a well known interior orientation and are stable. In this environement one major impediment to automation used to be the determination of approximations for pose as well as relative orientation (relative pose of two cameras). Before CUMULI the user had to manually select targets, and provide approximations for the pose. With this the typical measurment time of an image was several minutes. This made it impossible to automate hand-held camera systems.
The developement of a 4 point pose detennination algorithm in CUMULI together with coded targets and an 'orientation cross'reduced the time required for one image to seconds. A major issue for practical use '''as the reliability of the process. The first implementation showed that under certain conditions, if the coded targets are misidentified, the algorithms did not work. This was improved by increasing the number of coded targets on the 'orientation cross' and by using 'robust' algorithms. The success rate came to 99.9%. During the certification process of a machine control system, ''hich lmebic 'kveloped together with a consortium and Boeing, the algorithms were further enhanced and allowed tIx: certification of the system for controlling NC machines at Boeing.
A further step ''as to eliminate the 'orientation cross'. In order to do this the 'relative orientation' of images had to be computed with a direct solution. Based on an algorithm developed by CUMULI lmebic made a first implementation. After thorough industrial testing with a number of industrial clients the solution ''as introduced in mid 2001 for standard use. Currently this technology has been advanced to a point '~re in all applications we achieve a 100% success rate with 0% wrong orientations.
There ,,-as significant ,\'ork done for the orientation of cameras with non-point type features such as fu¥::S. This technol~.. has not found its introduction into the main stream market, less due to technical difficulties but the type of market environment we are working in.
Technology developed within CUMULI is currently in use in all Photogrammetry systems developed by lmetric, in the ICam metrology cameras, in the TIZ machine control system, and in the IScan surface
digitizing systems. These technologies have received the ' Technologie Standort Schweiz' award and the Silver Eagle award from Boeing.

Références bases de données (Anglais)

Swiss Database: Euro-DB of the
State Secretariat for Education and Research
Hallwylstrasse 4
CH-3003 Berne, Switzerland
Tel. [phone]
Swiss Project-Number: 96.0095