Automatic multi-style license plate detection using a biologically inspired classifier

In this paper, we propose a bio-inspired classifier that provides high classification accuracy even if number of training images is relatively small. The proposed classifier is applied to the Egyptian license plate detection problem. It can detect multiple license plates with different sizes and orientations even in complex environment. The proposed approach has two stages. Firstly, candidate plate regions are extracted using a sliding window and a weak classifier. Then another classifier with more features is used to classify the candidate regions into the appropriate plate style. This approach has been applied to the Egyptian License Plates with three different plate styles. The proposed approach achieved a success rate of 92% using only 40 images per category as training data.