Abstract: Confines of declining angles and definition in the emendation as well as the affection of plate frame, rivet, and space mark to the extraction are the key problems to license plate recognition (LPR). An improved Harris corner detection algorithm utilizing maximums and template-match (HMT) is proposed. With the algorithm, the inner-corners can be extracted simply and exactly, not affected by the declining angles. On the basis of the inner-corers, the plate is emendated without frame and the characters are located at the center of the image. Thus the characters can be extracted accurately by considering them as a whole. The experimentol results show efficiency and good performance of the algorithm in emendation and extraction.

Key words: license plate emendation, Harris operator, inner-corner, character extraction

摘要： 倾斜角度、边框清晰度影响着车牌的校正，边框、铆钉和间隔符等也影响字符的提取，该文提出了一种改进的Harris角点检测算法，该方法不受倾斜角度影响，直接定位车牌内角点，可将内角点作为校正基准点。校正后的车牌不存在边框，字符串位于图像的中心位置，可以根据车牌字符串的整体位置值提取字符。实验证明，该算法在车牌校正与字符提取方面是有效的。

关键词: 车牌校正, Harris算子, 内角点, 字符提取

CLC Number: 

• TP18