The Research On Methods Of Visual Detection For Visible Contaminant Particles In Ampoule Solutions

In this paper, researches on visual detection of visible contaminant particles in ampoulesolutions are conducted, and a visual detection algorithm which simulates the procedure ofmanual inspection is proposed to ensure the detection accuracy and detection speed. Theapplication of the algorithm in the intelligent inspection system has greatly improved theefficiency and has brought a lot of economical and social benefits.Firstly, the development and applications of machine vision technology are brieflyintroduced in this paper, and based on this point, the paper has elaborated the applicationstatus of visual technology utilized in current medical environment at home and abroad andthe significance of automatic inspection machine which is apt for our country; Subsequently,the application situation and environment of the visual inspection machine in the domesticand international areas are explored and researched;Secondly, the commonly used methods for small target detection are described and wedivide these methods into two categories of TBD and DBT according to the procedure of theirprocess. Then the common methods of DBT are introduced in details and experimental resultsare given. At last, by compared the effects between these experimental results,we describedthe advantages and disadvantages of these two kinds of methods.Thirdly，after detailed study of the commonly used methods, the detection method in theintelligent inspection machine is focused in this article, which is based on the regularmovement pattern of foreign particles. An algorithm,which is based on an improved framedifferential method and is combined with the adaptive threshold detection algorithm, isproposed. This paper has analyzed the characteristics of the target in details and thedifficulties in detection. As to the mechanical instability, the light interference outside thebottle wall, and other interference problems, an algorithm is proposed which is conduced inthe following order: firstly, by correcting offset to eliminate the impact of mechanicalinstability; then, setting the rational time interval between the image acquisition andcombining with the improved differential method to detect targets; the experimental resultshas been given and verified that the inspection algorithm is effective and practical;Then, the tracking algorithm is properly designed to ensure the detection accuracy afterthe detection result of foreign substances in the solution. The improved method of kalmanfilter is used to predicate and correct the tracking. By setting a reasonable time interval in theprocedure of image acquisition and establishing the movement equations of the suspicious moving target, the actual situations and parameters in the initial state of motion are setproperly and experimental results have been given which have demonstrated the effectivenessof the algorithm.Finally, the research platform for the algorithm is presented in details, including visualinspection system which consists of the conventional inspection system, optical module,mechanical components and electrical parts, etc.; then, the software system is introducedbriefly in this paper, including the system-control module and in operate-status module. Thenthe visual inspection parts are paid much focus, including the adjustment of inspectionparameter, the adjustments of camera parameter and so on.