Design And Implementation Of Vision-Based Pet Bottle Packaging Quality Automatic Inspection System

Modernization of the production pipeline for beverage facilities provideshigh conveniences. Each canned beverage production line can producethousands of bottles per hour drinks. But at the same production line ofcanned drinks will encounter in the canning process of substandard products,which include substandard and the liquid level and bottle cap defect situation.In order to ensure production of quality products, a thorough range of testbottles is needed. The traditional detection method is based on manuallysorting bottles, which takes up the heavy workload and lacks of stability.First, the design of the system is divided into hardware and softwaresystems and integrated systems from the perspective of software and hardwaredetection system in all functions of the modular system design. Second, thehardware system is divided into mechanical transmission system, imageacquisition system, signal transmission system and bottles choosing system.With building the hardware, the system implements functions of systematictransmission, image acquisition, signal transmission and choosing. Third, withanalysis of images of defective PET bottles, the defective PET bottles can bedivided into two cases, one of which is a cap defect; the other is the leveldefect. Cap defects can be divided into three cases-high cap cover, crookedcap and no cap. Level defect can be divided into high level and low-levelliquid situation. Square method-based detection algorithm and the capregional integration based on level detection algorithm to identify the PETbottles good the defects. Finally, we integrate the hardware and softwaresystems, and do online experiments with the integrated system; on-lineexperimental results validate our algorithm with low computationalcomplexity and precision.The proposed algorithm can be used for beverage packaging productionline quality inspection system, PET bottles, after the actual testing, testing more than99%accuracy rate, detection rate in the100ms/frame, can detectthe equivalent of36,000bottles/hour.