Liquid Impurity Detection System Based On Trajectory

With the development of machine vision,the automatic detection technology based on machine vision has been greatly developed in the aerospace,medical and industrial automation field.It has become an important research direction to detect impurities in transparent or translucent liquid.In a transparent or translucent liquid container production line,the traditional method of detection is to rely on the human eye to observe the liquid impurities in the case of the backlight baffle,and then the selected containers that contain the impurities will be removed.Compared with the traditional method,the method based on image processing for detecting impurities in the liquid can save the human resources,increase effectively and improve the level of automation of the entire production at the same time,the workers can also avoid the disease caused by the long time contacting light.The existing algorithms mainly aim at detecting and tracking the moving objects in a stationary liquid container,such as frame difference method,background subtraction method,mean shift method and so on.However,The liquid container also keep moving in the production line,the stationary target's position in image will change with the liquid container's position changes,and then the algorithm will still regard vessel target as the surface as moving targets;due to the limitation of the camera while capturing images,the size and gray distribution of target changes obviously.It means that shape and gray features of the target is not stable,which makes some tracking algorithm based on those features of target tracking is very not easy to capture.These reasons make the algorithm form a relatively large false alarm and missing alarm during the process of detecting the impurities,which greatly influence the accuracy of the detection.The main work of this thesis:(1)According to analyse collected impurity image in the static environment,after observing the centroid motion characteristics of impurities,we design the impurity detection algorithm based on position correlation which effectively avoids the defects of single frame detection.(2)Analysis on different type of the motion in the production line,the movement of static target on container surface is analyzed,and we verified several different correction methods,comparing the will corrected position with the results of moving target's position,we find out a method to distinguish between stationary and moving targets and exclude the interference of container's surface static target,so as to reduce the false alarm rate effectively.