road sign recognition groups in the world

euler.fd.cvut.cz
the road sign recognition groups in the world


Contents :
Brief history excursion
A list of reasearch groups and activities
Publications and references related to the road sign recognition
Basic documents

Brief history excursion

The first works related to the road sign recognition have been published (as far as I know) in Japan in 1984. The aim was to try various computer vision methods for the detection of the objects in outdoor scenes. At the begining of nineties, several groups interested in the subject emerged. We may found there solutions focused on particular road sign type together with more general systems. There were reported various approaches for the sign detection (employing of edges, colour segmentation, correlation etc.) in these works. The classification step has been - in most cases - solved by the use of neural network. I find the compilation of Lalonde and Li from CRIM (Canada) the most valuable information source describing systems before 1995.

Presently, there exist several groups, involved in the road sign recognition, in the world. Undoubtly, the most advanced system is the German TSR (Traffic Sign Recognition System) developed at the University Koblenz-Landau in cooperation with Daimler-Benz.

A list of reasearch groups and activities

Universita di Parma	Department of Informatics Bertozzi, Broggi, Fascioli ARGO - semi-autonomous car based on Linux. Several vision algorithms
Koblenz-Landau Univ.	Image recognition lab. Priese, Lakmann, Rehrmann, Lutz, Klieber, Jens robust road sign recog. system, modular conception, NN class. 1994 real time, autonomous vehicle, road tracking, obstacle detection, lane analysis, real time road signs recognition
Research Center Ulm, Daimler Benz AG	Ritter, Bartneck, Zheng, Yong-Jian, Werner, Janssen, Reinhard 1990 colour sequences, high order NNet, hypotheses, regions in the image 1992 colour segmentation, polynomial classifiers 1994 VITA II vehicle (hybrid parallel PowerPC), k-NN class., Radial Basis Function network (RBF) (for onchip implementation)
Texas A&M Univ	Dept. of Electr. Eng., College Station,TX, USA Kehtarnavaz, Estevez, Griswold, Kang 1994 feature extraction from occluded images based on colours, NNet 1996 TMS320C40, RGB segmentation, stop and yield signs, HSV segment., Sobel, Hough transform., morphology, boundary countour
Genoa Univ. Italy	Dipartimento do Fisica Piccoli, de Micheli, Parodia, Campani, Nicchiotti, Ottaviani, Castello gray level and colour (if possible), edges, cross-correlation with candidate region, Kalman filtering. classification of scene objects (cars, signs, buildings...) shape based and colour based approaches optical processing (spatial light modulators)
Inst.fur Algoritmen und Kognitive Syst. Karlsruhe	Frank, Haag, Kollnig 3D model for occluding scene situations
US Army Env. Fort Meade	Kellmayer, Zwahlen driver assistence or highway signs inventory system, BP NNet, fixed colour segmentation, warning signs
Dept.of Transport. Mng. Taipei, Taiwan	Fann Jeun-Haii, Lee Gang colour based signs finder for highways
Osaka Nat.Research Inst.	Matsuoka, Taniguchi, Mokuno Opticall correlations for road signs detection
ENST de Bretagne, Brest	Guibert, Attia Optical correlator for road signs recognition, PSA Peugeot Citroen
GRPR, Ecole Polytechnique, Brest	Cohen, Herve, Tong, Gouailler, Fourt, Lina. An Integrated System for Road Condition Diagnosis and Management (local html abstract)