基于机器视觉的经编机在线瑕疵检测技术

On-Line Vision-Based Fabric Inspection Algorithm

文中提出了一种离线学习、在线检测的经编机瑕疵检测方案。在离线情况下,利用小波参数方程构造完备的正交小波集,采用遗传算法寻优获取与待检布匹相匹配的最优小波;在线检测时,利用最优小波对待检图像进行单层分解,对瑕疵信息分布集中的高频分量进行子窗口分割,提取窗口内的灰度方差作为纹理特征描述,采用自适应双阈值分割方法自动检测瑕疵。实验表明,自适应小波方法对瑕疵信号有明显的增强作用,检测算法的准确率能够保持在98.5%以上。

A method of off-line learning and on-line detection of defects for warp knitting machine is proposed.In the off-line situations construction complete orthogonal wavelet set using parameter equation and optimized with genetic algorithm to obtain the optimal wavelet.The on-line measurement of detection using optimal wavelet to decompose image,and segmentation of high frequency component into windows,extraction the gray variance in the child window as the texture features,using the adaptive threshold segmentation method for the automatic detection of defects.Experiments show that the adaptive wavelet method of flaw signal,has obvious reinforcing effect,detection accuracy can be maintained at more than 98.5%.