Abstract: In order to solve the problems of the traditional straightening machines, such as the lack of adaptability, the dead angle of recognition, the poor precision etc., we propose a weft-straightening system based on camera detection. In the system, we use a fabric texture identification algorithm with the auto-registration of Fourier transform magnitude spectra. With threshold division based on statistics, we establish a mapping of spectra image on polar coordinates, so as to get the fabric-inclining parameter which is to be used in the auto-control system of the weft-straightening machine. Computer simulation demonstrate that the proposed algorithm can adapt to the identification of several different kinds of fabric as woven, stockinet, damask and mesh fabric.

Key words: camera detection, weft-straightening system, image processing, texture identification, Fourier transform

摘要： 光电整纬设备存在织物品种适应性不强、检测识别死角、检测精度不佳等缺点。本文提出基于摄像检测的整纬系统，使用CCD摄像机取代传统光电式整纬机的光电探头。采用傅氏变换功率谱自配准性质的纹理识别算法，利用基于统计的阈值分割在极坐标系上建立映射关系，从而得到织物的纬斜参数，再通过整纬器的自动控制系统对布料的纬斜程度及时做出纠正。通过对机织布、针织布、斜纹布和网眼布等不同种类织物的仿真验证了该算法的有效性和适应性。

关键词: 摄像检测, 整纬系统, 图像处理, 纹理识别, 傅氏变换