Abstract: The preprocess procedure of the automatic check processing system is produced, including the extraction and segmentation of the handwriting numerical strings. Proposed a practical method to extracting numerical strings: background and grid lines are first removed; then seals are segmented by threshold determined by objects size and deviation; finally, numerical strings are extracted through region analysis. A numerical string segment method is also proposed, which is based on grid lines. Experiment results demonstrate the effectiveness of the proposed method.

摘要： 设计了支票小写金额自动识别系统的前期处理模块,包括手写数字串的提取与分割。在数字串的提取过程中,首先采用动态阈值选取算法去除图像中的背景底纹,然后设计专门的线检测与去除过程去除定位格线;再利用检测到的框线位置将图像分块,根据目标的大小选取合适的阈值去除印章图像;最后采用连通区分析的方法提取数字串。在数字串的分割过程中,结合了定位格的位置信息,简化了分割 过程。该文提出的方法在现行支票图像上获得了较为满意的结果,基本达到了实用的水平。

关键词: 阈值, 文档图像分析, 数字串, 分割

CLC Number: 

• TP391.4