Car license plate recognition through Hausdorff distance technique

Abstract

Many researches have been done on car license plate (CLP) recognition. However, these are no sufficient number of works on Thai CLP. The existing works on Thai CLP have been focused on localization, extraction, and partial recognition (only for upper line) of the license plate. Generally, Thai CLP has 2 lines (upper and lower). The upper line presents character category in 2-position formatted, following by running number (up to 4 digits). The lower line presents province name. Recognition of the lower line is a difficult task because Thai words can be composed of 4 levels (top level, above level, base line and below level). Hence, appearance of noises in the original image easily causes misrecognition. This paper presents an approach to recognize off-line entire Thai CLP using Hausdorff distance technique (HD) for similarity measurement. The recognition rate obtained by the proposed method is 92%