基于最近邻算法的棉花异质纤维图像自动识别

Automatic Identification of Cotton Heterogeneous Fiber Image Basedon Neighborhood Algorithm

提出基于最近邻算法的棉花异质纤维图像自动识别技术,以提高纱线质量和杂质检测的准确性.在开卷机上,将棉绒加工成薄薄的棉层,并将采集的棉层、异纤、伪异纤图像保存到工控机上,针对异物目标区域,获得目标区域的颜色、形状和纹理特征.以棉纤维加工过程中存在的异物为对象,采用神经网络、最近邻算法等方式,对异物进行线性和径向基核函数支持向量机处理,并应用改进的粗糙集理论提取异物特征,得到有效特征向量.实验证明,所提方法提取的棉花异质纤维特征相对更加精准,能更好地实现棉花异质纤维识别的研究要求.

An automatic recognition technique of cotton heterogeneous fiber image based on nearest neighbor algorithm is proposed to improve yarn quality and impurity detection accuracy.On the uncoiler,the cotton wool is processed into a thin cotton layer,and the collected cotton layer,different fiber and pseudo fiber images are saved to the industrial control computer,the color,shape and texture features of the target area are obtained for the target area of foreign body.The object is the foreign body in the process of cotton fiber processing,using neural network,nearest neighbor algorithm and so on,the foreign body is processed by linear and radial basis kernel function support vector machine,and the improved rough set theory is used to extract the foreign body feature to obtain the effective feature vector.Most The experimental results show that the characteristics of cotton heterogeneous fiber extracted by the proposed method are more accurate and can better realize the research requirements of cotton heterogeneous fiber recognition.