Research On Key Technologies Of Vision-based Impurity Inspection System For Pharmaceutical Liquid

Due to the defects of manufacture and sealing technologies in the automaticproduction line, some foreign impurities occasionally appear in the pharmaceuticalliquid like infusion solution, ampoule injection, vial and so on. Once injected into theveins of patients, the impurities may lead to severe danger, even death. Actually, theinspection is currently implemented by trained inspection workers in most Chinesepharmaceutical corporations, which could not guarantee the safety of pharmaceuticalconsumption. This thesis focuses on the key technologies of vision-based impurityinspection system for pharmaceutical liquid, including vibration insensitivity of theinspection algorithm and diminishing vibration disturbance from theelectromechanical system. After a series of theoretical derivation, experiments,improvement and optimisation, an inspector prototype has been developed andapplied in the field successfully to replace manual inspection.The main contributions of this thesis include:1. The principle of machine vision based inspection technology for liquidpharmaceutical is discussed in detail. An overall inspection scheme is proposed. Thesystem parameters are calculated. The inter-frame difference algorithm for visibleimpurities inspection in liquid pharmaceutical is given. Finally, the prototype of thesingle position inspector for visible impurities in liquid pharmaceutical is developed,which is essential for sample testing, scheme confirming, and parameters calculatingwhile developing automatic online machine vision based inspector.2. An overall design scheme of automatic online machine vision based visibleimpurity inspector for liquid pharmaceutical is proposed, in which the servo trackingsystem is adopted to achieve online inspection. In addition, the control scheme ofphoto-mechatronic integration system and synchronous tracking curve are designed.3. A vibration-insensitive inter-frame difference inspection algorithm for onlineimpurities detection is proposed to diminish the vibration disturbance of automaticonline machine vision based visible impurity inspector for liquid pharmaceutical. Theinspection algorithm employs an efficient subpixel image registration method toperform a pre-registration of neighbouring images. A mountain-climbing searchingstrategy is proposed to improve its efficiency. Besides, a specialised computer clusterbased on gigabyte Ethernet is built to achieve real-time inspection. 4. The electromechanical system of the automatic online machine vision basedvisible impurity inspector for liquid pharmaceutical is optimised. An anti-impactwheel is added to diminish the vibration disturbance. A test software is developed toimprove the accuracy of the electromechanical system. After improvement andoptimisation, the automatic online machine vision based visible impurity inspector forliquid pharmaceutical well satisfies the designed accuracy and speed, and is applied inthe automated production line of the pharmaceutical manufacturers successfully.