In recent years, license plate recognition system has become a crucial role in the development of smart cities for vehicle management, investigation of stolen vehicles, and traffic monitoring and control. License plate recognition system has three stages, including license plate localization, character segmentation, and character recognition. Up to now the license plate recognition system has been successfully applied to the environment-controlled smart parking system, however it still raises many challenges in the surveillance system such as congested traffic with multiple plates, ambiguous signs and advertisements, tilting plates, as well as obscure images that are captured during bad weather and poor light conditions. In this thesis, we propose an efficient license plate recognition system that first detects vehicles and then retrieves license plates from the detected vehicles to reduce false positives on plate detection. Thereafter, the technique of convolution neural networks is applied to improve the character recognition accuracy from the blurred and obscure images. The experimental results show the superiority of the performance in the proposed method as compared to the traditional license plate recognition systems.