医用大输液杂质微粒分步检测系统研究

Research on Two-step Impurity Particle Detection System for Pharmaceutical Infusion

为提高塑料瓶装大输液中的杂质微粒的检出率,降低误检率,研究了一种分步式检测系统。该系统由圆盘检测台、仿人工检测台和交换工作台3部分组成,分‘旋转-急停’、‘仿人工翻转’两个步骤,结合超声波消泡技术、利用图像处理方法完成异物检测。通过设置‘仿人工翻转’机构将瓶身从竖立缓慢翻转至水平,以便于获得清晰的杂质微粒序列图像,同时减少机械随机抖动对图像质量的影响。对竖立和水平两种位置采集的序列图像进行分步处理与识别,竖立状态获取的图像处理区分静、动态目标,剔除瓶身皱褶、划痕和刻线等影响,标出动态可疑目标;水平图像处理可识别塑料碎屑微粒、复检其它可疑杂质。厂内模拟试验及高速自动生产线实际应用表明,与传统‘旋转-急停’检测系统相比,分步式检测系统的误检率和漏检率分别降低了0.55%和0.3%,对浅色异物检出率提高了27%,能够更好地满足医用大输液高速自动化生产过程中异物检测的需要。

In order to improve the detecting accuracy of impurity particles in plastic bottled medical infusion,a two-step detection system was developed.The system is composed of three parts:disc detection platform,human-imitation detection platform,and switching worktable.By using the image processing method,foreign bodies were detected based on the images acquired on the rotation-emergency stop mechanism and human-imitation detection mechanism.Ultrasonic technology was applied to defoam in the liquidI.tiseasy to get high quality and clear image sequence,and at the same time,to reduce the influence of random mechanical jitter by slowly flipping the bottle from vertical to horizontal on the human-imitation mechanism.In the disc detection platform,images acquired in the vertical position were used to distinguish static with dynamic targets,eliminate the disturbing of bottle wrinkles,scratches,lines,and marks.On the human-imitation detection platform,images acquired in the horizontal position were applied to identify plastic debris particles and re-examine other suspicious impurities.The in-plant simulation test and the practical application of the high-speed automatic production line show that,compared with the traditional "rotary-stop" detection system,the mistake rate and the miss rate of the two-step detection system is reduced by 0.55%and 0.3%respectively,and the detection rate for light color foreign is raised by 27%,which can better meet the needs of foreign body detection in the high-speed automatic production process of medical infusion.