Binary-channel can covers defects detection system based on machine vision

Abstract

In order to realize the on-line detection and elimination of unqualified can covers, a binary-channel inspection system based on machine vision is proposed in this article. The system mainly consists of two illumination sources, two cameras, two sensors, an IPC, an interface circuit, two eliminating devices, a set of algorithms for image processing and a software for the control of independent binary-channel. In working status, each channel of the system is placed directly above a conveyor, which transports can covers to the detection position so that the camera is triggered, and then an image is captured with a flash. The cover images are transferred to the IPC and then processed by the algorithm that based on template matching and variation model. Depending on the processing results unqualified covers are eliminated. The system is proved to be non-pollution, low-cost, and defects such as double covers, no glue, shoulder scratch, distortion can be detected with a 98.7% accuracy and a speed of 1200 covers per minute.