基于频率信息的织物起毛起球等级检测

Evaluation of Fabric Pilling Based on Frequency Information

针对织物毛球含有的信息和织物纹理结构等干扰信息频率不同的特点,采用Haar小波变换将图像按频率分层,依据纹理结构产生频率信息在数量上的优势,以及纹理信息分解的方向性和毛球信息分解的各向异性,找到纹理及毛球所在频率层。在毛球层上计算起毛起球密度,毛球大小,毛球形状等特性参数,进而依据这些参数实现未知织物类型的起毛起球等级判定,判定准确率达到91.3%。

The frequency of fabric pilling is different from the frequency of fabric texture. This paper uses the Haar Wavelet to decompose the fabric image in multiresolution, finds out the frequency level of texture and pilling based on the advantage in number and direction of texture frequency, and the disorder of fabric pilling, calcutaes the density, size, and shape of pilling. Based on these features, the pilling graded can be decided without knowing the fabric kind, and the accuracy can achieve 91.3%.