基于机器视觉的轻薄弹性裤袜破洞类疵点检测

Defect detection method for thin elastic pantyhose based on machine vision

为了甄别在弹性裤袜中可能会出现的损坏性破洞,提出了一种基于机器视觉的检测方法。利用工业摄像头获取裤袜图像,选用合适的小波基,对经过预处理的裤袜图像进行二维离散小波变换,从小波分解得到的裤袜经向和纬向子图像中提取特征值,对含有疵点的裤袜图像的高频细节子图进行特征提取和归一化处理。对得到的特征值曲线进行分析,根据曲线的突变位置,判断待检测裤袜中疵点的位置;采用基于最大熵的阈值分割法进行阈值分割,从而获取裤袜疵点位置。

A hole detects detection method for the thin elastic pantyhose based on machine vision was proposed. The pantyhose image was taken by industry camera. Selected a suitable wavelet base and transformed the preprocessed image of pantyhose by the two-dimensional discrete wavelet. Extracted and normalized the characteristics from the high-frequency detail sub-image. The position of defect of the pantyhose was determined by analyzing the characteristic value curve. Furthermore, maximum entropy was applied to threshold segmentation to locate defects.