Integrated Vision Units for Process Monitoring and Feedback in Electronic Assembly

This paper first briefly reviews the wide variety of machine vision (MV) tasks that are often performed in the assembly of electronics. Then the generic algorithms reported in the literature are cited for the most common and essential tasks. From this background, a range of required machine vision performance characteristics are suggested. These are couched in terms of resolution, pixel rates, and computing requirements. In order to meet these tasks an approach using highly parallel integrated vision units is suggested. “Integrated Vision Units,” IVUs, are individually of modest resolution and computing power. The word “integrated” refers to the construction of the individual vision systems. These systems are pure machine vision systems, where (1) the video detector, usually a Charge Coupled Device (CCD), is directly attached to a micro-processor without traditional camera circuitry, and (2) the individual units DO NOT support a human or video interface. This approach greatly reduces the cost of the individual IVUs and provides a level of performance not normally available in terms of illumination timing, pixel stability, and control of image acquisition (Lee, 1994). The argument is made that most of the electronic assembly related tasks can be handled by economical combinations of IVUs. For circuit inspection and measurements during and after assembly at typical production rates up to 50 IVUs in parallel might be used. Such a system would not cost more than a single high resolution camera system, yet would have pixel rates and computing power, 10x, such systems. For monitoring of individual components during placement, 10 IVUs in parallel would be a more realistic construction. For tracking of a fiducial, a single IVU can be used.