Dynamic Morphology-based License Plate for Color and Infrared Images

This paper proposes to use the method of edge detection, morphology, and dynamic image thickening for license plate extraction from images. In the proposed algorithm, a different thickening is used for rear and front parts of the image; besides, to increase the segmentation rate, determination of the license plate frame using standard deviation in the vertical histogram diagram is suggested. Furthermore, separation of the numbers and characters, which are stuck together due to dirt, is performed using the aspect ratio of the components and location of the valleys in the vertical histogram diagram in binary scale. The proposed method is a complement for license plate extraction, which is capable of detecting several license plates in the image. Also, by modifying the proposed method and highlighting the objects’ edges, it can be also used on infrared images. Results of the tests performed on a set of color and infrared images indicate improvement in the location rate and segmentation rate of license plates compared to the current methods.