摘要
探讨基于纹理特征的棉亚麻纤维识别技术。利用微分干涉相衬显微镜采集到能够清楚表征纤维纹理特征的图像,通过图像预处理,采用中轴线方法获取棉亚麻纤维直径和直径不匀率的形态特征,分别提取不同角度的灰度共生矩阵能量、惯性矩、相关性、熵等四个特征向量的均值和标准差,采取支持向量机,通过交叉验证法训练得到最佳参数对棉亚麻纤维进行分类识别,采用形态特征和纹理特征相结合的识别正确率达到94.4%。认为:采用形态特征和纹理特征相结合的方法能够提高纤维识别的正确率。
Identification technology of cotton flax fiber based on textural feature was discussed. Differential interference contrast microscopy was adopted to gather images that can clearly represent the fiber textural fea- ture. Through image pretreatment,axle method was used to obtain the morphological characteristics like cotton flax fiber diameter and diameter irregularity. The average and standard deviation of different angles gray-level coocurrence matrix four feature vectors as energy, moments of inertia, correlation and entropy were extracted sep- arately. Support vector machine was adopted. The optimal parameters for the classification and identification of cotton flax fiber were obtained through the training of cross validation. The identification accuracy by using the combination of morphological characteristics and textural features was up to 94.4G. It is considered that the method of adopting the combination of morphological characteristics and textural features can improve fibers iden- tification accuracy.
出处
《棉纺织技术》
CAS
北大核心
2016年第4期1-5,共5页
Cotton Textile Technology
基金
湖北工业大学高层次人才启动基金项目(BSD2012002)
关键词
棉
亚麻纤维
微分干涉相衬显微镜
灰度共生矩阵
支持向量机
纹理特征
纤维
直径
Cotton,Flax Fiber,Differential Interference Contrast Microscopy,Gray-level Cooccurrence Ma- trix, Support Vector Machine, Textural Feature, Fiber Diameter