摘要
金属表面具有高反光和拉丝特性,金属表面细小瑕疵的图像检测一直是非常困难的问题,本文提出了一种基于支持向量机的金属表面细小瑕疵检测算法。算法采用将图像分成小块并用支持向量机分类的方法。将图像分割为小的子块,适用于细小瑕疵的检测;运用机器识别的方法,克服了传统图像算法难以检测金属表面瑕疵的缺点。对近200个样本的测试结果表明该算法的准确率高达93.75%。
The metallic surface has the property of？？？high light and hairline, so the fine flaw detection on the metallic surface has always been a difficult problem. In this paper, an algorithm for defect detection based on support vector machine is proposed. In this algorithm, the image is divided into little blocks and classified by the support vector machine. The method of dividing the image into little blocks is fit for the detection of fine flaw , and by using the machine recognition method , the problem in the traditional image processing method can be solved. The test of 160 samples demonstrates this method is feasible and effective.
出处
《微计算机信息》
北大核心
2008年第33期191-192,184,共3页
Control & Automation
