Visual Inspection for Circular-Tablet Quality by Image Template Matching

This research develops a visual inspection system of circular-tablet quality according to three-dimensional tablet shape using image processing and analysis techniques. Multiple circular tablets are conveyed in a single row along the center-line conveyor belt into a light-control box so that image clarity and shadow minimization can be achieved. As multiple tablets passing 2 high-speed cameras, the tablets' height and diameter can be correspondingly inspected/measured in real time by top and side cameras. If tablets' shape in both cameras does not contain any defect, determined by template matching algorithm in OpenCV-python, then either single or multiple rectangular frames is drawn around each fine-quality tablet locations as well as tablets' volumes are computed. According to 100-sampled experimental tests, an average volumetric error is about than 11.85%. In cases where the tablet is defective: chipped or non-circular shape, the inspection program will not draw a rectangular frame around those tablets and the program will report to users to remove those defected tablets from the conveyor belt.