Character Recognition in Car License Plates Based on Principal Componentsand Neu ral ProcessingAline da Rocha Gesualdi, Jos´e Manoel de SeixasRio de JaneiroFederalUniversitySignal Processing Laboratory (LPS), Eletric Engineering Departament, COPPE/UFRJRio de Janeiro - RJ, Brazil, CP 68504, 21945-970aline@lps.ufrj.br; seixas@lps.ufrj.brMarcelo Portes de Albuquerque, M´arcioPortesde AlbuquerqueBrazilian Center for Research in Physics (CBPF)Technical Activity Department (CAT)Rio de Janeiro, RJ, Brazilmarcelo@cbpf.br, mpa@cbpf.brAbstractWe present, in details, the Principal Component Analysis(PCA) applied on a neural based recognition system. Thetechnique is valuated on the extraction of principal compo-nents in character images of Brazilians’ license plates. Thispaper focus on the usage of the PCA in the recognition ofcharacter images. Comparisons with different neural clas-sifi ers are made.1. Introdu ctionThe design of intelligent transportation systems for ap-plications such as electronic toll, traffi c fl ow analysis, speedlimit and red light violation enforcement has been attractedan increasing attention from researchers. In particular, therecognition of characters in car license plates is object ofdifferent studies nowadays. As a very powerful techniquefor pattern recognition problems, it is not surprising thatneural networks became the basis of a number of such sys-tem designs [8, 2, 1].Recently, a successful design for Brazilian car plates hasbeen reported [3].The design approach, which is sum-marized here for reference, included a number of prepro-cessing steps aiming at extracting effi ciently the charactersegments from digital image plates. Splitting the classifi -cation task into two, two expert neural classifi ers [5] weredesi...
