Abstract
This paper investigates knowledge transfer (KT) on Saudi License Plates (SLP) enhancement. Particularly, we investigate two models of KT, Representational and Functional. We use such models to transfer knowledge from the source neural network that was trained on recognizing the Arabic alphabet and Indian numerals in the old SLP style. We utilize transferred knowledge to increase target network performance and recognize additional numerals and alphabet systems in the recent SLP style. We implement, test, and compare such models against a conventional multi-layer network model. Simulation results show that KT models outperform the said model.