Damaged Licensed Number Plate Detection Based on Novel Object Oriented Classification with K-Means Algorithm

Abstract

The aim is to detect the quality of number plates using object oriented classification in comparison with K-Means clustering. Two groups such as novel Object Oriented Classification and K-Means algorithm are applied. Total number of samples that are evaluated on this proposed methodology are 265 images. Among this sample dataset, 185 images [70%] of the dataset was taken as a training dataset and 80 [30%] was taken as a testing dataset. Programming experiment was carried out for N=7 and N=9 iterations for novel Object Oriented Classification and K-Means algorithm respectively. Computation processes were executed and verified for exactness. SPSS was used for predicting significance value of the dataset considering G-Power value as 80%. Novel Object Oriented Classification algorithm shows a high accuracy and homogeneity for damaged number plate detection, and has recognition rate of 0.593 (p>0.05). This research article is intended to implement an innovative approach to Automatic License Plate Recognition for detection of damage in licensed number plates. Comparison results show that efficiency of Novel Object Oriented Classification is better than K-means Algorithm for detecting damaged number plates.