Inspection meets high-volume manufacturing

Sophisticated electronics for automotive mirror systems present a considerable challenge to their manufacturers. Features include compass and temperature indicators, telematics (communications), electrochromic (self-dimming) capability, and map lighting. At Magna Donnelly Electronics (Holly, MI; http://www.magnadon.com), a producer of automotive mirror systems, we needed to improve our inspection techniques to maintain product quality and to accommodate increasing production volumes.

Fig. 1. These automotive mirrors include vacuum fluorescent displays (VFDs) that provide direction and temperature information.           Fig 2. Data Translation evaluated this Compass product to design the best possible inspection solution.               Fig. 3. The new inspection system uses an optical character reader (OCR) algorithm to translate inspection images into alphanumeric information for pass/fail determination.

Each VFD consists of 42 independent elements. The modular system could handle up to 80-pixel windows per camera program. Because inspecting each VFD element required a dedicated pixel window, inspecting each mirror unit required six camera programs to achieve a 48-pixel window-inspection capacity.

Each inspection sequence involved the following steps:

• acquire an image;
• search for a particular detail within the image;
• determine the detail's x-y coordinates; and
• track peripheral pixel windows with those coordinates.

The modular inspection system returned a pixel value based on a preprogrammed color representation of an active VFD element, then repeated the process for all VFD segments. A full inspection took 7 s. The tester spent nearly 6 s of that time controlling the vision system and receiving pixel-window data, partly because an RS-232 serial interface linked the tester's PC to the vision system. We programmed the system through a menu on a remote-control console. We also developed a PC program to process the unit's serial data output.

Although the modular systems proved adequate at first, they lacked the flexibility and speed to cope with ever-higher production volumes. In addition, modified camera programs were difficult to back up and store because of their size (usually more than 2 Mbytes) and the speed (or lack of it) of serial data transfer. Backups could take as long as 15 min, time lost to production.

To overcome these drawbacks, I started looking for a suitable alternative—a PC-based vision system. Because Magna Donnelly's PC-based test systems run under Microsoft Visual Basic, I wanted an inspection system that would integrate easily into that environment.

A PC-based system offered several benefits: increased inspection speed, a greater selection of examination tools, and easier program version control. It could also save images acquired in production testing and transfer them to the PC, something the modular system couldn't do. Such a system would also allow me to select frame grabbers, cameras, and other hardware as test requirements dictate rather than limiting my choices, and I could develop complete inspection programs across a single software platform.

I contacted Data Translation (http://www.datx.com) for help. The Data Translation application engineer requested a sample PCB for evaluation. I supplied a product with a temperature display (Figure 2) configured to display the three production-test modes: all VFD segments active, alternate-segment test mode, and normal operation.

Data Translation suggested a system that featured a similar inspection sequence to that of the modular system—that is, identify a search detail, inspect at the detail's x-y coordinates, and return an RGB value for each VFD element. The new inspection system, which included Data Translation's Vision Foundry software, also allowed us to develop an optical character recognition (OCR) algorithm and perform a complete inspection with one camera program.

The new system required a learning curve—our engineers had to understand ActiveX controls and Distributed Component Object Model (DCOM) technology in order to create transparent data exchange between the test system's program and the remote server. Data Translation's technical support team guided us through the transition with phone support and several Visual Basic code examples. The migration took four to six weeks. Some of that time was spent exploring the inspection tools and some was spent with the application engineer learning the DT Vision Foundry DTiX information (image/data) exchange server and how to interface with DCOM.

Our new system performs inspections under one "point and click" script developed with DT Vision Foundry. The time to complete a vision acquisition and subsequent VFD inspection has dropped from 7 s to barely 1 s. The modular system returned a pixel count for each inspection pixel window and then compared those results to individual test limits for each camera program. The PC-based system can pass information from the acquired images directly to the Visual Basic application. An OCR algorithm simplifies and reduces governing test criteria. Rather than comparing pixel information on 42 pixel windows to determine VFD output, the new system simply translates the inspection images into an alphanumeric result such as a temperature value or a compass direction (Figure 3).

Our new system also offers programming-structure flexibility. The modular system offered about a dozen tools and could store only 16 programs. Each program supported a maximum of eight pixel windows, or a maximum of 128 pixel inspection windows altogether. An inspection script on the new system can incorporate nearly three dozen inspection tools in any order. If necessary, a single script can analyze thousands of regions of interest in addition to controlling other tools.

The new vision system's software generates script files that, to date, remain smaller than 1 Mbyte and reside on the PC's hard drive, so backup times become trivial. Microsoft Source Safe (MSS) serves as a version control tool for all test-application source code and support files. By using MSS to archive and track changes to script files, we have eliminated problems related to camera-program file maintenance. Version control, traceability, and archiving are now accomplished by exchanging script files with the MSS utility. The new system, which is flexible, easy to use, and easy to maintain, has greatly improved our inspection process.