Vehicle plate segmentation (VPS) is key technology in the vehicle plate recognition (VPR). There are numbers corners which distribute intensively and regularly in the vehicle plate region. An approach utilizing the corner distribute feature to segment the vehicle plate accurately and fast is presented. Detecting corners, finding out candidate vehicle plates and segmenting the vehicle plate accurately are three main steps of the approach. In the course of corner-detections, a great number of non-corner pixels are eliminated by using reasonable grade threshold, which largely reduce the expense of vehicle plate segmentation. The candidates are determined with a slid window and clustering analyses are used to segment the vehicle plate accurately. Experiments show that the approach is effective and fast. For the covering abstract see ITRD E134653.
Abstract