通过对倾斜的DR图像进行图像增强、二值化、倾斜校正、去毛刺、填补空洞、单字符分割和大小归一化处理,分割出若干个大小相同的单字符图像；然后通过训练45个支持向量机(SVM),并得到各个支持向量机的判别函数；利用这45个判别函数判别单字符图像是哪个数字的图像,并记下各个数字的得票数,确定得票最多的数字即为单字符图像对应的数字.应用Microsoft Visual Studi0 2008软件编写了相关的计算程序.利用该程序进行识别实验的结果表明,该方法能够在DR图像中货车铸件工件号呈倾斜情况下有效识别出货车铸件的工件号.%Through image enhancement, binaryzation, tilt correction, deburring, filling the empty, single-character segmentation and size normalization processing, slanted DR image is divided into several single-character images which have identical size. By training 45 Support Vector Machines (SVM), the discriminant function of each SVM is obtained. These 45 discriminant functions are used to judge which digital image the single-character image matches with, and note down all digital votes. The single-character image corresponds to the number is determined by the number of most votes. Microsoft Visual Studio 2008 software is applied to compile the relevant calculation program. The program is used to carry through recognition experiment. Results show that this method can effectively recognize the casting workpiece characters for railway freight train among slanted DR images.
展开▼
