摘要
以炼钢物料的自动识别为应用背景,提出了基于快速提升小波变换和支持向量机(SVM)的识别方法。该方法首先运用DB4小波的提升算法对图像进行"塔式"分解,提取小波系数统计量作为图像的纹理特征组成特征向量,利用SVM算法进行分类。在炼钢厂原料图片的分类实验中,该方法的分类准确率为99.15%,平均图像特征提取时间为0.074秒。实验结果表明,该方法已满足企业生产的要求,并且准确率和实时性优于该类应用的其它方法。
In steel mills, many steelmaking raw materials need to be identified automatically. A new classification method based on fast lifting wavelet transform and support vector machine (SVM) is presented. First of all, DB4 wavelet decomposes the material image in "tower" manner and then some wavelet domain sl:atistics are abstracted as the image facture vector, at last, images are classified via SVM classification algorithm. The experiment reaches a validity proportion of 99.15%; the average time for image feature extraction is 0.074s. The results show that this method meet the requirements of production, and accuracy and real-time are superior to other methods in such application.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第18期4093-4096,共4页
Computer Engineering and Design
关键词
炼钢物料
分类识别
提升小波
纹理
支持向量机
steelmaking material
classification and recognition
lifting wavelet
texture
SVM