Research On Liquor Foreign Body Online Detection Technology Based On Machine Vision

Health wine, as an alcoholic beverage with health care function, with the growingawareness of people’s health is gradually favored by consumers. However, in the filling andcapping process,it’s inevitably to produce or mixed foreign bodies such as cullet, flocculent,hair. The existence of this foreign matters have a great impact on the quality of the products,especially the glass debris will endanger consumers’ lives. The detection equipment based onmachine vision which is an emerging technology of automatic detection is widely used inforeign country and its price is expensive, while the traditional backward artificial lightinspection method is still used in domestic factory. In order to promote domestic wine visibleforeign matter detection technology forward, the visible foreign matter detection, tracking andrecognition technology is studied and the overall scheme foreign object detection system isdesigned.Based on the statistical classification of the visible foreign matter, modules such as theseparation modules of foreign object and background, lighting modules and removingmodules are designed, and the final hardware architecture and system software are completed,providing guidance for the follow-up work.Algorithm of the foreign object detection based on image sequences is studied. As thetarget is small, an improved weighted median filtering algorithm for noise filtering isproposed. On the basis of the classic two adjacent frames difference method, an improvedthree-frame difference method is proposed, and the experimental results show that thismethod can effectively detect foreign objects, enhance the foreign body gray and is conducivesubsequent tracking, processing recognition algorithm.To track and recognize the visible foreign matters, the kinematic of high density matter isanalyses, and the result shows that the trajectory in the vertical direction is exponentially.Tracking the moving object based on the Mean Shift algorithm. An recognition algorithmbased on support vector machine which is characterized with the moving target position,shape and trajectory is proposed...