基于边缘检测的邮件地址标签图像分割方法

Method of Locating Address Blocks in Parcel Post Images Based on Edge Detection

为了提高邮件分拣的效率,邮政分拣中心投入使用了自动分拣设备。尽管OCR系统对印刷体字符的识别已经达到了很高的准确率,但是邮件地址标签的准确定位问题仍然是分拣系统设计中的难点。一般情况下邮件尺寸不一、地址标签位置不固定、标签表面经常有污损,这使得地址标签的定位变得非常复杂。在经典图像处理理论和大量实践的基础上,提出了一种基于边缘检测的图像分割方法———行扫描灰度渐变法。该方法在邮件自动分拣的图像预处理环节中取得了良好的效果,经实践证明是一种可行的图像分割方法。

Postal sorting automation has become imperative with the large volume of mail that must be handled everyday. Although OCR systems are available that can read machine print with great accuracy, the region of the parcel post containing the destination address has first to be found and locating, it proves to be the major impediment to automatic mail sorting. There is generally no fixed position for the address block on the parcel so that the task is made more complicated. An efficient method is presented for automatically locating address blocks in images of parcel post. The image is segmented into address blocks based on size, position, and pixel density. The success of the method has been demonstrated for several test images.