一种医用口罩缺陷的机器视觉在线检测系统

An Online Medical Mask Defect Detection System Based on Machine Vision

针对口罩生产中口罩片短小、鼻梁条缺失或过短、耳绳缺失、断裂或短小、油污、破洞等8类缺陷,设计一种基于机器视觉的口罩缺陷在线检测系统。硬件方面,设计出的高亮均匀的多光源光学系统具备颜色自适应性,且能直接照射出口罩内层污点,极大降低了后期检测算法的复杂度;算法方面,首先对几何校正和定位后的口罩图像采用改进自适应Canny算子进行边缘检测,再根据各缺陷特征完成口罩片、鼻梁条和耳绳缺陷检测,并提出了一种基于Blob分析的油污、破洞缺陷检测算法。实验结果表明,算法检测精度达0.3 mm,平均检出率为99.1%,口罩检测速度高达120 pcs/min,可满足医用口罩实时在线生产需求。

Aiming at 8 kinds of defects in the production of masks,such as short mask body,missing or short nose bar,missing,breaking or short ear rope,oil stain and hole,etc,an online mask defects detection system based on machine vision is designed in this paper.In terms of hardware,the high-bright and uniform multi-light source optical system of color adaptability can directly illuminate the inner stain of the mask is designed,which can greatly reduce the complexity of the later detection algorithm.In terms of algorithm,firstly,the improved self-adaptive Canny operator is used to detect the edge of the mask image after geometric correction and location,then a Blob analysis detection algorithm of oil pollution and hole defect is proposed.Then the defects of the mask body,nose bar and ear rope were detected according to different defect characteristics.Experimental results showed that the detection algorithm’s accuracy was up to 0.3 mm,the average detection rate is 99.1%,and the detection speed is as high as 120 pcs/min,which can meet the real-time online requirements of medical mask's production.