图像五值化与基于三值图像的车牌识别.doc

图像五值化与基于三值图像的车牌识别( 张忠义1+ 1(海口愚佬会教育科技有限公司, 海南 海口 570203) Image 5-Value-Segmentation and License Plate Recognition based on 3-value image* ZHANG Zhong-Yi1+ 1(Haikou Yulaohui Education Technology Co., Ltd, Haikou 570203, China) + Corresponding author: Phn: +86-898 Fax: +86-898 E-mail: yulaohui@ Received; Accepted Abstract: Image 5-value-segmentation is a image simplification method used to map RGB space into YB space, by the formulation Y=(R+G)/2 and then convert color in YB space to five values: black、blue、gray、yellow and white. For Chinese license plate recognition, five values can further be compressed to three values by combining black and blue, white and yellow together separately. The method proposed has two application scenarios: character recognition based on 3-value image and license plate location by inversing segmentation information based on 3-value image. In the first scenario, The reserved gray in 3-value image can be classified into either the plate background color or the foreground character color for the character recognition according to the license plate background color recognized. It can make up for the information loss in the process of image binarization by determining threshold, which is always influenced by the uneven illumination or dirty spots. Doing this helps to improve the accuracy of character recognition. In the second scenario, candidate location of license plate can be found in the premise of traversing the image efficiently if character segmentation can be performed based on projections of the 3-value image. This method can detect character spacing firstly to meet the needs of users’ and eliminate false license plate，make sure the accuracy of license plate location which doesn’t depend on license plate character; it can also reduce the true negative by ignoring the inference brought by plate features. Key words: 5-value segmentation; license plate recognition 摘 要: 图像五值化是通过公式Y ＝（R＋G）／2，将RGB空间映射到YB空间，再将YB空间简化成黑、蓝、