Skip to Main Content
License plate localization is the most important part of the license plate recognition system. Ability to correctly detect license plate under different conditions directly affects overall recognition system accuracy. In this paper a real-time license plate localization method is proposed. First, vertical edges are detected from the image and binarized. Then, license plate candidates are extracted by the two-stage detection process. In this process a sliding-window technique is used to mark all windows which satisfied edge density conditions. Edge density conditions are computed on integral edge image allowing us to significantly increase the processing speed of the method. To better distinguish between license plates and complex backgrounds, the edge analysis is performed to remove specific edges. Finally, false candidates are filtered out based on geometrical and textural properties. The proposed method can detect multiple license plates with different sizes in a complex background. The experimental results confirm robustness and ability to localize license plates in real-time. On the database of 501 images our method correctly localizes 97.4% of license plates.