摘要
针对棉/亚麻混纺织物,基于其单纤维纵向显微图像(纤维切段的长度约为0.5 mm),研究了纤维的自动识别方法。检测纤维时,先对纤维图像进行去背景处理,而后运用形态学闭运算和背景区域生长相结合的方法获得纤维的目标区域,对图片中出现的玻璃划痕、干扰杂物等进行了较好的滤除。由纤维骨架垂直方向上的区域图、二值图和细化图得到它们的垂直积分投影序列,并提取这3条序列各自的变异系数CV值和平均值共计6个参数。将这6个参数作为棉/亚麻纤维的特征参数,训练最小二乘支持向量机分类器,对测试集的测试结果表明该分类器对棉/亚麻短纤维的识别正确率平均为93.3%。
Aiming at cotton/flax blended fabrics,a new automatic identification method based on the longitudinal view of microscopic fiber images is proposed,in which the length of fiber about 0.5 mm is used for image capture.For fiber detection,the background of fiber image is removed firstly,then fiber areas are detected by a method combining morphological close operation and background regional growth,and the glass scratches and other sundries in the images are filtered as well.Based on the region image,binary image and refining image of binary image perpendicular to the fiber skeleton,their vertical integral projection series are captured,and each coefficient variation（CV value） and mean value of these three series are extracted and used as the texture parameters of cotton/flax blended fabric to train the least square support vector machine classifier.The experiment results show that the mean identification accuracy of cotton and flax fibers is 93.3%.
出处
《纺织学报》
EI
CAS
CSCD
北大核心
2012年第4期12-18,共7页
Journal of Textile Research
关键词
棉纤维
亚麻纤维
纵向切段
混纺比
检测
图像识别
特征提取
支持向量机
cotton fiber
flax fiber
longitudinal view
blending ratio
detection
image recognition
feature extraction
support vector machine
