Bhutanese License Plate Recognition Using Hu’s Moments and Centroid Difference

The purpose of this study was to develop an automatic number plate recognition system for Bhutanese license plates. This paper presents a license plate localization based on state-of-the-art YOLO (You-Only-Look-Once) object detector. Once the region of interest was extracted, the connected component analysis method was proposed with some additional method to segment the characters and handle different types of license plates. The features from the segmented characters were extracted using Hu’s moments, and the Centroid Difference (CD) method was proposed to handle characters having the same geometrical shape. The extracted features were trained and tested on WEKA using random forest classifier algorithm and the obtained accuracy was 94.6%.