A Hybrid KNN-SVM Model for Iranian License Plate Recognition

Abstract

This study presents a new method for Iranian License plate recognition systems that will increase the accuracy and decrease the costs of the recognition phase of these systems. In this regard, ahybrid of the k-Nearest Neighbors algorithmand the Multi-Class Support Vector Machines (KNN-SVM) model was developedin the study. K-NN was used as the first classification model as it is simple, robust against noisy data set and effective fora large data set. The confusion among the license plate similar characters problem was overcome by using the multiple SVMs classification model. The SVMs model has improved the performance of the K-NN in the recognition of similar characters. The current study experimental results revealed that there is a significant improvement in the character recognition phase rate compared with a similar study.