基于SURF特征的羊绒羊毛鉴别算法

Identification algorithm of wool and cashmere based on SURF feature

为了实现羊绒与羊毛纤维的快速自动鉴别,提出了基于SURF特征的羊毛羊绒鉴别算法。首先采用扫描电子显微镜获取两种纤维的表面形貌图像,再对图像进行预处理,去除背景并对纤维区域进行增强处理,然后提取SURF特征并聚类,将原始图像转换成相对应的高维向量,最后对数据进行分类。试验结果表明,该方法有效,在两种纤维不同混合比例的情况下,识别率都超过90%。

To achieve the quick and automatic identification of wool and cashmere, a practical algorithm based on speed-up robust features（SURF） is proposed. Firstly, the surface images of the two fibers are taken by scanning electron microscope（SEM） , and the image is preprocessed, the background is removed and the fiber region is enhanced. Then the SURF features are extracted and clustered, and the original images are converted into corresponding high-dimensions vector data. Finally, the data is categorized by support vector machine （SVM）. The results show that the proposed algorithm is an effective approach for identification of wool and cashmere with average recognition rates over 90% with different mixing ratios of wool and cashmere.