摘要
为了提高工业字符识别的准确率,增强字符识别算法对含噪声字符或发生形变字符的适应性,提出了一种改进的轮廓层次特征提取方法。对经过预处理归一化的字符,先提取轮廓层次特征,再对特征信号进行小波分解,从分解结果的低频部分中提取特征信息,最后将特征输入SVM(Support Vector Machines,支持向量机)训练和分类。实验结果表明,该特征提取方法降低了后续要处理的数据量,具有良好的抗干扰能力,实用价值较高。
In order to improve the accuracy of industrial character recognition and enhance the adaptability of our character recognition algorithm to noisy or deformed characters,a method of improved contour feature extraction is presented.First,contour features are extracted form the normalized characters,and then are decomposed by wavelet transform.Character features are extracted form the low-frequency part of the result of the decomposition.In the end,the SVM classifier is trained by sample characters and then is used to recognize test characters.Characters on industrial components are choosen as recognize objects.Experiments show that the presented method reduces the feature dimensions and improves the anti-interference ability of the features which have higher practical value.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第6期2423-2427,共5页
Computer Engineering and Design
基金
国家自然基金项目(60804013)
"嵌入式机器视觉关键技术的研究与开发"产学研合作基金项目(011002)
关键词
字符识别
特征提取
轮廓层次
小波变换
支持向量机
character recognition
feature extraction
contour feature
wavelet transform
SVM
