期刊名称：International Journal of Signal Processing, Image Processing and Pattern Recognition
印刷版ISSN：2005-4254
出版年度：2015
卷号：8
期号：9
页码：363-370
DOI：10.14257/ijsip.2015.8.9.39
出版社：SERSC
摘要：It is difficult to track, count and separate the moving bars at a high speed on production line for their overlap and accumulation. Therefore, it is necessary to establish a reliable, practical recognition and segmentation mechanism for the adhered bars. A new solution to the problem of bars adhesion is proposed: a support vector machine is constructed to recognize the adhesion type of bars by the feature vectors of training samples. The geometric feature values and moment feature values based on Blob regions in images are extracted, which is the input feature vector of support vector machine. The trained classifier is used for identifying the adhesion type of bars in images. Finally, classification and recognition is carried by support vector machine. The experimental results show that the recognition accuracy based RBF kernel achieves 100%. The method is feasible and effective for the recognition and segmentation of the adhered bars.
关键词：adhering bars; support vector machine; classification; recognition; bars
