Automatic license plate detection based on edge density and color model

This paper proposes a novel method for license plate (LP) detection from images with complex background. First, it segments images with an adaptive binarization method to avoid the problem that nonuniform illumination creates, and some undesired image areas are removed by limiting the range of region properties of connected components (CCs). Secondly, CC analysis is used to construct nearest neighbor chain (NNC) for detection of candidate LP regions (LP-NNC). The average height and direction of each LP-NNC is estimated to deal with images acquired from different view or distances. Thirdly, length of NNC, edge density and color features are combined to verify all candidate LP regions, and the most possible region is selected as the true LP region. Experiment results on various types of LP images show that this proposed method has achieved desired detection result for complex scenes.