摘要
为了解决字符识别算法在噪声、文本旋转下出现识别不准确的问题。论文提出了基于二维Gabor滤波与机器视觉的字符识别算法,从图像预处理和特征提取分析这两个模块展开分析。首先,搭建视觉采集硬件平台,采集原始字符图像。然后通过邻域像素灰度信息迭代,有效降低噪声,以提高字符图像质量,通过霍夫直线检测完成图像旋转角度计算和图像校正,达到准确定位字符区域的目的。然后,通过二维Gabor滤波得到字符图像的纹理特征,采用欧式距离构建分类标准,达到准确识别字符的目的。实验测试数据表明:与当前识别机制相比,在面对噪声与文本旋转干扰条件下,论文算法具有更高的识别准确率与鲁棒性。
In order to solve the character recognition algorithm in noise,identify inaccurate problem under text rotation. Char-acter recognition algorithm based on computer vision is proposed in this paper,from the image preprocessing and feature extractionand analysis of the two analysis module. First of all,visual acquisition hardware platform is created,the original character image iscollected. Then through iterative neighborhood information noise reduction the quality of character image is improved,by hoffstraight line detection the image rotation angle calculation and correction are completed to achieve the purpose of accurate position-ing character area. Then,by 2 D Gabor filter to get the texture feature of character image,the classification criteria of Euclidean dis-tance was used to construct and achieve the purpose of accurate identification character. Experimental test data show that comparedwith the traditional recognition mechanism,this mechanism has higher recognition and robustness.
出处
《计算机与数字工程》
2017年第12期2519-2523,共5页
Computer & Digital Engineering