【作者】 张斌；

【导师】 毛建旭；

【作者基本信息】 湖南大学 ， 电子科学与技术， 2015， 硕士

【摘要】 医药玻璃瓶大输液是我国医药行业五大重要制剂之一,是医疗机构日常经常使用的药品,在现代临床上占据十分重要的地位。但是,由于生产设备及检测设备的局限,在灌装过程中,大输液产品中可能含有玻璃屑、纤维、毛发等可见异物,这些不溶性异物进入人体血液后会危及到人的生命安全。目前,我国制药企业基本上都是采用人工灯检的方法。这种落后的检测方式存在很多缺陷,如检测速度慢、精度低,易受干扰,可靠性差,还容易对药品造成二次污染等,无法从根本上保证产品质量,也无法满足高速生产线的生产要求。首先,介绍了玻璃瓶大输液视觉检测系统的研究背景和意义,结合大输液药品的生产流程和机器视觉技术在工业检测领域的应用,分析了医药大输液视觉检测技术与设备的国内外研究现状,然后根据大输液中可见异物的特点,提出了玻璃瓶大输液中可见异物视觉检测系统的整体设计方案,系统详细介绍了光源和照明方式,工业相机,机电控制系统和系统检测软件架构等关键技术。其次,对获得运动异物的连续图像进行预处理,其主要目的是为了抑制成像时背景噪声的干扰。针对干扰噪声,本文提出了一种自适应中值滤波算法,与传统的滤波算法比较,该算法能够去除噪声的同时较好地保留边缘等细节信息,降低图像处理后的模糊化程度,对检测区域内的药液图像进行了噪声滤除。然后,当检测系统运行时,系统机械振动不可避免,工业相机和大输液瓶体之间不能保证绝对同步,因此连续获取的大输液序列图像之间会出现偏差,必须对获取的连续帧图像进行配准处理,才能准确地检测出可见异物,本文提出了基于二进小波与SIFT特征的图像配准算法。该算法将原始图像和模板图像进行二进小波分解以获得粗尺度的平滑图像;再利用Do G特征点检测算子对图像进行关键点检测,进而采用欧氏距离对特征点进行匹配。实验证明,所提算法不但可以有效减少图像特征点匹配的计算量,而且提高了图像特征点的匹配率,从而增强了图像配准算法的实用性和高效性。研究并实现了基于帧间二次差分与灰度能量积累相结合的异物目标检测方法,有效的排除了瓶身上可能存在的各种痕迹干扰,准确检测出大输液中的异物目标。最后,在研究以上技术的理论基础上,开发出了一套大输液中可见异物视觉检测软件。设计了用户信息模块、控制系统模块和软件检测模块,并详细介绍了软件系统中各功能模块的开发与实现过程。更多还原

【Abstract】 Pharmaceutical glass bottle large transfusion is one of the five important preparations pharmaceutical industry in China. They are frequently-used drugs in medical institutions and play a very important role in modern clinic. However, some visible foreign substances may appear in the bottle infusion uring the processes of producing, filling and packaging. These foreign substances will turn out to be a serious threat to patients. At present, most of the Chinese pharmaceutical manufacturers adopt the traditional manua l inspection. Obviously, the manual inspection is slow, complex and unreliable, and usually brings in extra pollution to the drugs, thus is unsafe, Can’t guarantee the quality of the product fundamentally, but also can’t meet the production requirements of high-speed production line.Firstly, the background and significance of the glass bottle large transfusion visual detection system research has been proposed in this paper. Combined with the application of infusion drugs production process and machine vis ion technology in the field of industrial detection, analysis the research status of medical infusion visual detection technology and equipment at home and abroad, and then put forward the overall design of glass bottle large transfusion visible foreign ma tter in the visual detection system according to the characteristics of visible foreign substances in transfusion system are introduced in detail light source and lighting, camera industry, mechanical and electrical control system and detection system software architecture and key technologies.Secondly, the continuous images which obtain moving foreign bodies are pretreated, and its main purpose is for the interference of the background noise suppression imaging. According to the interference noise, this paper presents a kind of adaptive median filter algorithm, with traditional algorithms, the algorithm can remove noise and preserve edges and details better and reduce image processing after the fuzzy degree, can be effective in the region detection liquid image noise filtering.Thirdly, when the detection system run, system of mechanical vibration is inevitable, between industrial camera and the transfusion bottle can ’t guarantee absolute synchronization, thus continuously acquires the transfusion between sequential images will appear deviation, must be on the acquisition of the consecutive frames image registration processing, can accurately detect visible foreign matter, this paper proposes based on the dyadic wavelet and scale invariant feature transform(SIFT image registration algorithm. In this algorithm, the original image and the template image of dyadic wavelet decomposition to obtain the coarse scale image smoothing; reuse thedog feature point detector for detecting the key points of image. Then the Euclidean distance of the feature points matching. Experimental results show that the proposed algorithm can not only reduce the calculation amount of image feature points matching, but also improve the image feature points matching rate, thus enhancing the image registration algorithm is practical and effective. Research and implementation of inter secondary difference points and gray energy accumulation combination of foreign objects detection method based on, effective discharged various trace interfer ences that may exist in the bottle, accurately detect the infusion of foreign objects.Finally, a large infusion of visible particles in visual detection software is developed based on the above technology. Design the user information module, control module and software detection module, and introduces the development and realization process of each function module of system software.更多还原