【作者】 张哲；

【导师】 朱铮涛；

【作者基本信息】 广东工业大学 ， 仪器仪表工程（专业学位）， 2014， 硕士

【摘要】 近年来,药品安全一直是政府治理、媒体报道以及社会舆论的重点关注问题。于是,与药品有着密切关系的胶囊也成为社会关注的焦点。一些胶囊生产企业对胶囊的检测已经利用胶囊检测设备并通过随后的人工抽检,然而这依然无法满足产业化生产的检测需求和检测效率,从而严重制约着企业的发展。随着计算机技术的快速发展,带动着机器视觉系统和数字图像处理的发展。机器视觉系统是用机器视觉检测方法代替人工视觉检查产品质量效率低且精度不高的不足；数字图像处理是指将图像信号转换成数字信号并利用计算机对其进行处理的过程。基于这样的背景,检测系统以其客观、高效、精确、非接触的特点,成为视觉系统中重要的技术手段。国产胶囊检测机的出现,彻底改变了原有胶囊依赖人工检测的现状,提高了胶囊产品质量,有效地防止可能出现的二次污染。但是,由于国内胶囊检测机机械很多是从胶囊印字机改造而来,先天的不足,导致了检测结果不太理想,诸如不能有效展开表面,漏检率高,检测速度低而检测机成本高等,针对目前国内胶囊检测设备存在的问题,本文重点研究提升胶囊检测系统性能的若干关键问题。针对胶囊样品的特点,提出黑白和彩色相机的多特征、多相机信息融合机制完成胶囊图像采集的方案,利用多个不同位置布置的黑白和彩色相机,实现胶囊全表面的成像,研究了光电传感器和编码器触发采图方案,最大限度了获取展开表面的图像并减少CPU的负荷；针对图像数据量大及对在线快速处理的要求,研究胶囊图像快速处理的方案,利用Halcon函数库中函数的并行运算机制,计算机的多线程技术以及图像处理计算策略,提高运算速度；针对快速在线准确剔除废品的运用需求,研究了基于电磁铁和电磁阀的剔除方案,设计了专用的晶体管通断实现剔除的控制电路。本文的特色在于利用多种策略保证基于胶囊图像实现缺陷快速性、稳定性和精确性的检测要求。更多还原

【Abstract】 In recent years, pharmacy security has always been a key problem that government t rying to deal with and has drawn attention to the public and media. Capsule, of course, re mains a focus relating to pharmacy security. Human vision inspection is applied to inspe ct capsule detects after using some inspection machines, but these techniques are unable to meet the needs of capsule manufacturing requirements. Company is restricted in devel opment correspondingly. Machine vision and image processing technology can not only boost inspection efficiency and accuracy, but also replace human vision inspection. Imag e processing involves converting image signal into digital signal that computer takes in a nd calculate the data, a significant technique because it’s much objective, highly-efficien t, accurate and non-contacting.The arising of domestic inspection machine has totally changed human inspection a nd of course capsule quality also improved. Potential second contamination is avoided. B ut these domestic machines are genetically deficient and usually transformed from the ca psule printing machine, undesirable results are seen such as bad surface unfolding, high omission ratio, low speed and high cost, etc. The study mainly researches providing key solutions to improve the property of capsule inspection machine.The study has proposed multiple features of black-white, colored camera, informati on fusion of multiple cameras to complete capsule image capturing. Capsule surface ima ging using photoelectric sensor and encoder trigger scheme is researched by studying the black-white, colored cameras of different locations, which can ultimately obtain surface image and reduce computer memory consumption. As to large size of image data and on-line multi-thread operation, study has used image processing library Halcon to run parall el computation. Multi-thread operation and strategy of image processing have improved speed. Defects kicking scheme is researched using electromagnet and magnetic valve, so on-ff control circuit has been designed t meet the needs of on-line kicking. Key feature i n the study is that multiple strategies ensure the fastness, stability, and precision of capsu le inspection.更多还原