Part Metrology and Defect Detection Using Machine Vision

Abstract

Abstract This paper focuses on the detection, classification & sorting of good and defective blade/automotive fuses through the utilization of Machine Vision Inspection (MVI). Fuses are an electrical safety device found in most vehicles and serve to protect their wiring and electrical devices against overcurrent. These small-size components are produced in large quantities, through a process involving the encasing of two metal prongs connected by a thin metal strip inside a plastic housing. Due to the nature of working with delicate metals and plastic, during the manufacturing process multiple defects can occur. The purpose of the project was to detect, classify and sort good and defective fuses, while performing part metrology solely on good fuses. Using NI Vision Builder Software, an image acquisition algorithm was written to generate a database of images for all the fuse classes. With this database, further algorithms were constructed to identify each class of fuse as it progressed along the conveyor. Utilizing vision inspection techniques such as pattern matching and distance measurement, the system was able to properly classify the fuses in the database. If the system determined the fuse to be defective, an I/O signal from the smart camera activated a pneumatic system that would remove the fuse. However, when the system classified a good fuse, it was allowed to continue along the conveyor and be collected after part metrology was performed by another algorithm. A graphical user interface (GUI) displayed these inspection statistics in realtime. This project was successful in completing all the defined goals for the system and proved to have an object classification accuracy of over 90%. This MVI system has the proper algorithms to be used in tandem with a high-speed smart camera, for adaptation into a real-world manufacturing application.