Machine vision automates glass machine-tool inspection
Since feature dimensions need to be traced back to the specific tool that formed them, measurements must be conducted on the forming machine itself, directly downstream from the tool.
Glass vials are manufactured from glass tubing, which is processed by forming tools to shape the neck, lip, thread, and other glass features. Since feature dimensions need to be traced back to the specific tool that formed them, measurements must be conducted on the forming machine itself, directly downstream from the tool. As each glass vial moves around the perimeter of the forming machine, it also rotates on its own axis, enabling a series of measurements to be made around its circumference.
To automate this measurement process, Cynosure (Perkasie, PA), a specialist in gauging systems for the glass industry, has developed a machine-vision system that monitors the production stream for out-of-tolerance parts. This system is also linked to a reject mechanism that controls the forming process so that machine-tool malfunctions can be rapidly analyzed without delay or wasted production.
To obtain the required details, the gauging system captures eight frames of data in approximately 120 ms. For this application, Cynosure uses a 6600 dual-channel analog camera from Cohu (San Diego, CA) that operates at 60 frames/s and interfaces to an IC-RGB frame grabber from Coreco Imaging (Billerica, MA) and its MVTools software.
Based on a Pentium processor with MMX technology, the custom interface software was developed in Microsoft's Visual Basic 5. This software permits the technician to calibrate the optical system, set product tolerances, and continuously monitor the gauging process via a graphical user interface. Running under Windows NT in a network link, it manages statistical analysis off-line.
Data capture begins with a contact closure on the forming machine, signaling that the target is in the camera's field of view. After capturing the first eight frames, the frame grabber delivers a reset strobe to the camera, which collects subsequent frame sets as separate units. The frames are then transferred to the frame grabber for capturing and reformatting before being passed to the host memory for processing.
At this point, vision algorithms developed under MVTools analyze the video frame data and extract glass-vial dimension details. Initially, glass-vial parameters, such as top and side dimensions, are determined from a series of 60 or 70 caliper measurements. Every frame is then analyzed against 35 preset dimensional criteria to characterize the vial. Dimensional measurement data are then logged for presentation on the technician's computer screen.