基于受控形态学的无限制票据分类

Unrestricted Check Type Identification Based on Regulated Morphology

目前的票据类型识别要求票据中必须存在框线、分类码等明确的分类标志,而现有的实际票据往往不满足这些条件,基于受控形态学操作的二次分类方法选取图像本身的结构特征进行了分级受控击中击不中变换,对票据样式没有任何限制。分析和实验表明,该方法计算复杂度低、速度快、准确率高。

At present, the identification of check types need some marks on the checks such as frame lines or classified code. However, the classified marks are not shown in all the actual checks. An identification approach of twice classification based on regulated morphological operations which is transformed hierarchically with regulated hit-or-miss transformation by the structural feature of the image, is proposed without any restriction to the check style. Analysis and experiments show that the approach is with low computational complexity, high speed and high correct rate.