医药注射剂中可见异物在线视觉检测方法及关键技术研究

项目名称： 医药注射剂中可见异物在线视觉检测方法及关键技术研究

项目编号： No.61305019

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 葛继

作者单位： 江西理工大学

项目金额： 26万元

中文摘要： 药品是国家的战略物资，其质量更涉及国家安全、关乎国计民生、社会稳定和经济发展。然而，国内 99%以上的制药企业对药品出厂前的关键质量检测仍然采用人工检测方法，完全依赖肉眼观察，误检、漏检率高，难以满足药品生产质量管理规范与自动化生产需求。本项目以医药生产线为背景，研究适合国内制药标准的注射剂中可见异物视觉检测系统，重点突破药液中微弱异物目标的运动机理、最优视觉成像方案选择、高速亚像素级药液图像配准、微弱目标PCNN分割以及基于OS-ELM模型的实时多特征异物分类识别等系列处理算法，构建高速医药自动化生产线上注射剂中可见异物的实时视觉检测方法体系，并将研究成果应用于实际注射剂生产线上进行验证。项目所取得的理论与方法将有效解决药液中微弱可见异物视觉检测的瓶颈难题，大幅提升检测的精度和速度，保障药品质量安全。相关技术还可推广至食品、饮料、电子等质量检测领域，具有重大的科学研究价值和社会经济意义。

中文关键词： 安瓿注射剂；可见异物检测；机器视觉；视觉检测机器人；极限学习机

英文摘要： Medicine is the strategic assets of the nation. Its quality concerns national security, social stability, national economy and the people's livelihood. However, 99% of pharmaceutical corporations in China employ the manual visual inspection method to check their products. Omission error rate synchronously increased due to inspectors using their eyes for long periods of time. This method can hardly satisfy the Good Manufacturing Practice(GMP) and automatic manufacturing. Based on the high-speed automatic medical line, the vision-based foreign particle inspection system within injections will be researched trying to fit our medical clauses effectively. Faint foreign objects' moving mechanism, selection of the illumination styles and medical images' serial processing algorithms like high-speed sub-pixel registration, PCNN model based object segmentation and OS-ELM algorithm based real-time multi-feature particle classification will be broken through intensively. Real-time vision-based foreign particle inspection within injections can be structured and these research findings will be verified on the real production line. Bottleneck problems of the faint foreign particle detection shall be resolved based on the research results. The inspection accuracy and speed can be improved significantly and medicine quality can

英文关键词： Ampoule Injection；Foreign Particle Inspection；Machine Vision；Vision-Based Inspection System；Extreme Learning Machine