Research on Real-Time Quality Inspection of PET Bottle Caps

摘要

Online quality inspection is an effective way to guarantee the quality of PET bottles packaging on the high-speed filling production line. This paper focused on studying and developing a machine vision inspecting system for on-line defects detection of PET caps without support rings. An image processing algorithm was proposed to detecting the serious defects such as surface foreign matters and link ring broken of the PET caps. For detecting the defect of surface foreign matters, the algorithm employed a mean filtering to estimate the background, then compared the inspected image with the background image by using dynamic threshold, as a result, the crack region was extracted. As to defect of link ring broken, the uneven gray-scale distribution in the RGB color space was extremely improved by HSV color space transform. Compared three components of HSV images, the gap feature between the cap and the link ring was more obvious in the saturation image. Hence, the gap region was extracted through segmenting the saturation image. Experiments demonstrate that the designed inspection system has an accuracy up to 97% and the speed up to 34000 bottles per hour.