Using Mean Shift for Iranian License Plate Detection

محورهای موضوعی : Communication

Mohtaram Zandian ¹ (Islamic Azad University, South Tehran Branch)
Sedigheh Ghofrani ² (Islamic Azad University Tehran South Branch)

تاریخ دریافت : 1397/04/24 تاریخ پذیرش : 1397/05/29 تاریخ انتشار : 1397/09/10

کلید واژه: Cluster analysis, Mean Shift, Iranian License Plate Detection, Kernel Density Estimator,

چکیده مقاله :

In this paper, Mean Shift (MS) as a clustering algorithm is used to localize the Iranian license plate. In this procedure after clustering, based on the optimized MS method, we applied the geometrical features and edge density in order to remove those parts in every cluster, which cannot be the license plate. The main advantages of MS are no need to know the number of clusters and it is completely independent of the Iranian license plate characters colors or background colors. However, for MS implementation, we should only predetermine a parameter named bandwidth. The experimental results show that our proposed method achieves appropriate performance. We should mention that our system accuracy for optical (OP) with 300 images is 94.6% and for infrared (IR) with 80 images is 98.3%.