Tesseract视觉耦合支持向量机的字符识别算法

Character Recognition Algorithm Based on Tesseract and Support Vector Machine

为了解决当前字符识别算法严重依赖固定匹配模板,使其难以识别形态多变的手写字符的问题,本文提出了Tesseract视觉耦合支持向量机的字符识别算法。首先,通过开源视觉库Tesseract自带图像预处理工具,对字符图像进行边缘检测处理,提取字符的边缘特征;再通过训练工具cowboxer,快速训练出字符特征文件,通过识别函数与定位函数,准确完成字符的初步识别。然后,引入支持向量机,通过训练字符特征向量,开发训练字库,对初步识别中的遗漏目标完成字符的补偿识别,有效确保字符的识别正确率。实验结果显示:与当前识别算法相比,本文算法的识别精度与抗干扰性更高。

In order to solve the defect of difficult recognizing the morphological changing handwritten characters induced by relying heavily on the fixed template in current character recognition algorithm, the character recognition algorithm based on Tesseract visual coupling support vector machine was proposed in this paper. First of all, the edge feature was extracted by detecting the character image edge based on open source vision library Tesseract comes with image preprocessing tool; then accurate recognition of characters was finished by identifying function and the position function based on training tool cowboxer to fast train out of character feature files, the compensation recognition of the initial recognition with the omission of the target was finished by training the character feature to development of character training based on support vector machine. Experimental results show that this algorithm had higher identification precision and anti-jamming.