Publication details

Road sing classification using Laplace kernel classifier

Journal Article

Paclík Pavel, Novovičová Jana, Pudil Pavel, Somol Petr


serial: Pattern Recognition Letters vol.21, p. 1165-1173

research: AV0Z1075907

project(s): VS96063, MŠMT, IAA2075608, GA AV, IAA2075606, GA AV

abstract (eng):

The Laplace kernel rule for the road sign classification based on a priori information about road signs grouping has been developed. The smoothing parameters of the Laplace kernel are optimized by the pseudo-likelihood cross-validation method using the Expectation-Maximization algorithm. The new classification algorithm has been successfully tested on more than 1100 images of 43 road sign types. The comparison with the Bayes classifier assuming the Gaussian mixtures has been made.

Cosati: 12B, 09K

RIV: BB