一种基于数学形态学的汽车牌照提取算法

定位汽车牌照在车牌识别应用中是很关键的一步。作者提出一种基于数学形态学的新方法,以车牌宽度和高度,车牌字符高度及字间宽距等信息为依据来设计一种新的结构元,通过对汽车边缘图像进行形态学运算,能从图片中得到包含汽车牌照的侯选区域,最后基于汽车牌照纵宽比等固有特征,采用连通域体态分析,对包含车牌的多个侯选区域进行去伪,得到真正车牌区域。通过对大量汽车图片实验,验证了该方法的有效性。

An Approach Based on Morphology for Car License Plate Extraction

Locating the car license plate in a car image is an important step in car license plate recognition applications. This paper presents a new method based on morphology for license plate extraction from car images, which is combined width and height of car license plate, height of character, and space between characters of the car license plate. New structural elements for morphological operations are designed with the above features, and used to the edge image of car license plate, which can produce several candidate areas. According to the intrinsic feature of height to width ratio for a car license plate, connected component analysis is performed and the true car license plate area is isolated. The experimental results with a reasonably large set of car images are very encouraging.