中文摘要：

根据车牌字符的固有特征，提出一种新的基于投影的车牌字符分割方法。该方法首先对车牌图像进行预处理，检测车牌倾斜角度，如果倾斜角大于指定角度则进行车牌倾斜校正，然后利用车牌的水平方向投影去除车牌的上下边框以及铆钉，对处理得到的图像进行二值化。再根据车牌字符的排列规则和字符间距的关系，利用车牌的垂直投影定位字符，先分割出第二个和第三个字符，从第三个字符开始分割出后五个字符，再利用已分割字符的知识来分割前两个字符，然后对分割出来的候选字符块进行处理，有效解决字符粘连和断裂的情况，最终实现车牌字符的准确分割。实验结果证明，该方法有较好的分割效果。

英文摘要：

According to the intrinsic characteristics of license plate, a new approach for characters segmentation of license plate based on projection is proposed. Firstly, some preprocesses are processed toward the license plate images - detect the incline angle of license plate and rectify the slanted and distorted plate if the incline angle is bigger than the designated angle, then the horizontal boundaries are removed by using horizontal projection of license plate. After these processes, image binarization is processed to the image. Then the characters are located by using vertical projection of license plate, according to the ranging rulers and intercharacter distance of license plate characters. The second and third character are first segmented, then the last five characters from the third character are also segmented. With the knowledge of single character which has been segmented, the first two characters are segmented. Then all the possible characters are processed specially, segmenting the conglutinant characters and combining the cracked characters if existing. As a result, all the characters are segmented accurately. The experimental result shows that this approach has a good effect of segmentation.