Skip to Main Content
Considering the fact that High Definition (HD) becomes an important trend in road surveillance video, this paper studies vehicle location and license plate location methods in HD surveillance video. While license plate reading may obviously benefit from high definition technology, higher resolution also increases the computational load of graphical analysis and background interference. Most known approaches to license plate location are not suited to high definition imagery. In this article, a real-time method of license plate location over high-definition surveillance video is discussed, and a reasonable approach to consider HD is proposed. As license plates are affixed on vehicles, a prior vehicle detection step significantly enhances the robustness of license plate detection. It is shown that the frontal area of the vehicle can be located using an AdaBoost cascade classifier. Following this classification step, the region of Vehicle License Plates (VLPs) can be located based on fixed color and texture features of license plate characters and background in HSV color space. This paper also presents extensive experiments using thousands of real video sequences to verify the proposed method.