基于机器视觉的PET瓶口快速缺陷检测

Quick Defect Detection of PET Bottle Mouth Based on Machine Vision

为提高PET瓶口缺陷检测精度和检测速度,提出了一种基于机器视觉的PET瓶口自模板快速缺陷检测法。首先在全局阈值分割的基础上,对ROI区域(瓶口端面)构造自模板圆环,并且确保自模板图像和阈值分割图像坐标的对应关系不变。然后在ROI区域上,将自模板图像和阈值分割图像灰度做差进行缺陷检测,避免图像定位和缺陷定位对检测带来的大量复杂计算。通过实验验证,该方法缺陷检测精度达到99.9%,检测时间在50ms以内,有效提高了PET瓶口缺陷检测精度和速度,可以满足工业生产线上对PET瓶口缺陷检测准确性和实时性的要求。

In order to improve the accuracy and speed of PET bottle mouth defect detection, a fast PET bottle mouth self-template defect detection method based on machine vision was proposed. First, on the basis of global threshold segmentation, a self-template ring is constructed for the ROI area(the end face of the bottle mouth), and the corresponding relationship of coordinates between the self-template image and the threshold segmentation image is ensured to be unchanged. Then in the ROI area, defect detection is performed by subtracting the self-template image and threshold segmentation image, avoiding complex calculations brought by image positioning and defect positioning. The experimental results show that this method has a defect detection accuracy of 99.9% and the detection time is within 50 ms, which effectively improves the accuracy and speed of PET bottle mouth defect detection and can meet the requirements of the accuracy and real-time PET bottle mouth defect detection on the industrial production line.