Automatic License Plate Recognition (ALPR) systems have many applications in today’s traffic monitoring and toll-gate systems. License plate localization is the most important part of the ALPR system, because it is the most challenging stage. In this paper, we propose a real-time algorithm for localization of license plates in video sequences. Unlike methods with high computational complexity, we employ simple and effective techniques to be realtime.First, frames containing moving objects are obtained by using Gaussian Mixture Model (GMM). Then, candidate regionsare found by vertical edge detection and horizontal projection. After that, license plates are localized and extracted by morphological operations and connected components analysis.The proposed method is evaluated on 498 frames containing plates in different conditions from highway cameras. Our real-timesystem detects nearly all of the vehicles plates in video sequences and correctly localizes 98.79% of the license plates in these frames.It can also detect multiple plates in each frame. This system is implemented in C++ using OpenCV library and for HD video, 1280x720 pixels, the average processing time of 25 ms per frame was achieved. It means 40 fps which can be used in real time applications. Experimental results confirm robustness of the method.