Naslov
Video image speed limit sign detection in various conditions using neural networks

Autori
Fištrek, Tomislav ; Lončarić, Sven

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
Proceedings of the IADIS International Conferences: Computer Graphics, Visualization, Computer Vision and Image Processing 2010 / Yingcai Xiao, Roberto Muffoleto and Tomaz Amon - Freiburg, 2010, 433-438

Skup
MULTI CONFERENCE ON COMPUTER SCIENCE AND INFORMATION SYSTEMS - Computer Graphics, Visualization, Computer Vision and Image Processing

Mjesto i datum
Freiburg, Njemačka, 27.07.2010. - 29.07.2010

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
image analysis; pattern recognition; traffic sign detection

Sažetak
In this paper, we present a method for traffic sign detection based on an artificial neural network. The system extracts features necessary for the detection of a road sign. Searching of the picture is performed by scanning, and by extracting various features. Colour is taken as an important feature. That is a part of the spectrum in which the detection in the picture is the most successful for the given sample. The size of the searching frame is somewhat larger than the expected road sign size, however within the frame the net is trained to be invariant to scaling and position. On the basis of the combination of these features the neural network decides if a road sign is detected or not. Experimental results included several neural networks that are compared according to the speed and quality of the road sign detection and position in the image. The influence of the different choice of features and their combinations is investigated as well. Experimental validation has shown promising results.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo

POVEZANOST RADA