Research And Development On Online Vision Detection For Small Foreign Particles In Filling Liquid Medicine

China’s annual filling liquid medicine production and sales is the largest in the world, but there are more than 90% of the pharmaceutical manufacturers using artificial light inspection method to detect the visible foreign particles, in fact. Due to the key techniques restriction, domestic automatic foreign particles inspection machine does not work well. Therefore, it has great economic benefit and very strong practical value to develop online automatic filling liquid foreign particles detection system.This paper takes 10 ml ampoule bottle filled medicine liquid as the main research object. On the basis of detailed analysis of the existing technical, this paper takes the technology of machine vision and image processing to design a kind of image detection software, which is suitable for the filling liquid production line. And this paper uses the hardware such as PLC, servo motor, industry camera and lighting to build the experiment platform.For bottled liquid micro foreign particle visual inspection system, due to the platform and camera vibration, resulting in the bottle in image sequences with varying degrees of random offset, this paper proposes a liquid medicine image registration method based on modified phase correlation. First, this paper introduces the projected image into origin phase correlation registration algorithm to divide it into two one-dimensional registration algorithms. That can greatly improve the algorithm’s operation speed. Second, this paper introduces Hanning window function filtering to overcome the harm to the registration accuracy due to the “discontinuity” of the image edges. Third, this paper introduces the peak estimation technique to improve the registration accuracy to sub-pixel level.Second time difference and energy accumulation is introduced to solve the traditional frame differential method cannot clear difference results in the image target belongs to frame and weakening of the difference after target gray value problems. In this paper, a target tracking method based on differential structure similar is provided. First, this paper sets the moving step optimization to avoid the iterative process falling into a local optimal value. Second, this paper introduces the Kalman filter to predict the starting point. Third, this paper adjusts the error matrix of Kalman filter adaptively to improve the accuracy of prediction. The experimental results show that the proposed method can effectively track the fast-moving objects or shelter, background disturbance in severe cases. And it is very suitable for filling liquid visible foreign body extraction.Through theoretical analysis and repeated experiment tests, this paper designed the sequential control of the servo motor and developed camera control module and image acquisition module based on the JAI camera SDK development package. This paper built the platform software for the detection system using multi-threading technology and MFC framework.Finally, this paper successfully designed single bit prototype experimental equipment, based on the above efforts. And test the scheme of the system on the experimental equipment.