A robust license plate detection algorithm based on multi-features

This paper presents a robust vehicle license plate detection algorithm based on multi-features, including mathematical morphology, rectangle features, edge statistics and characters features. Here we utilize the word “robust” to describe the proposed algorithm because it is not only adaptive to variance occasions, such as variance of the illumination, vehicle position, the color, both background and foreground, the acclivitous angle and the size while working in different complex environment, but also can be used in several regional license plates while some other algorithms just work well for one region and badly for another. We have used the algorithm for detecting both American and Chinese plates and gotten a good performance in both regions. Those are the two primary contributions of the algorithm. Another contribution is that we use a character feature verification algorithm to determine the final detection result in the candidate rectangles. This is proved to be more effective than other features.