License plate recognition is an indispensable part in the management system of city intelligence traffic and the location of license plate is a critical step in license plate recognition. The traditional recognition algorithms of license plate mainly adopt edge transformation to locate license plate, and the degree of accuracy is not high. In this paper, an improved precise location algorithm of license plate is put forward. The picture taken by a camera is usually a color image, so the grayscale transformation is firstly discussed. Then, the gray image is converted to a binary image and the image binaryzation' critical step is how to find the optimum threshold. To eliminate the background interference and use vertical projection method to find the candidate region of license plate is the next step. Then, using median filter algorithm eliminates image noise. Lastly, the horizontal region of license plate is found out by horizontal projection and using vertical location acquires the precise location of license plate. The experimental results show that this algorithm discussed in this paper can precisely locate the license plate region and effectively enhance the recognition rate of license plate.