Vision system checks every display pixel
Vision system checks every display pixel
Andrew Wilson
Cellular telephones, medical monitoring equipment, and automobile indicators are a few of the many applications for alphanumeric and liquid-crystal displays (LCDs). Before such products are shipped, however, the displays must be operationally checked at the development, subassembly, and final manufacturing stages. To perform display tests effectively, display manufacturers, systems integrators, and in-house manufacturers are harnessing the power of PC-based machine-vision systems that integrate off-the-shelf lighting, optics, cameras, frame grabbers, and software-development tools.
"In telecommunications, especially," says Doug Wilson, general manager at the Engineering Technology Center (ETC) of Analysis & Technology Inc. (Mystic, CT), "systems developers are concerned about the quality of their products." Automated vision-based test systems enable manufacturers to meet increasingly more-stringent quality standards with repeatable, objective testing at a substantially lower cost than using subjective human-operator tests.
Wilson and his colleagues have developed DisplayCheck, a PC-based vision system configured to test the integrity of LCD and electroluminescent (EL) displays and the user-interface firmware. Designing the software to handle a variety of display illuminations proved to be a major challenge for ETC engineers. For the fastest and most accurate inspection, the vision system incorporates a fiberoptic ring light from Fostec (Auburn, NY) to illuminate the display under test. In some finished product applications, such as cell phones, integrated curved lenses between the inspection system and display dictate the use of the product`s own backlighting, which generally varies greatly over the display area. DisplayCheck includes robust dynamic, adaptive, background-normalization, and thresholding algorithms that eliminate the errors associated with poor lighting conditions.
"External lighting generally results in glare from the display glass," says Ken Pietrzak, the principal engineer on the project, "making it difficult to obtain a high-fidelity image." To overcome this problem, Pietrzak incorporated a polarizing filter in front of the lens of the RS-170 camera used in the system. "Because LCDs polarize the reflected light, a polarizing filter is aligned to transmit only that polarity of light," says Pietrzak, "and results in a high-fidelity image of the display with no glare."
To capture the image, the system incorporates an IMAQ-1408 PCI-based frame grabber from National Instruments (Austin, TX) that allows images to be displayed in real time on an SVGA or XVGA monitor. Initial versions of the system have been developed to test displays as large as 192 ¥ 120 ¥ 8-bit resolution.
"In such systems," says Wilson, "a minimum 4:1 oversampling (camera pixels to LCD pixels) using the 768 x 480 pixels of the RS-170 signal allows each pixel in the display up to 192 x 120 pixels to be inspected individually. For even higher-resolution displays, DisplayCheck can be configured with a digital area- or linescan camera and a National Instruments IMAQ-1424 digital frame grabber. "Having a flexible architecture allows us to customize the system to a particular set of requirements, thus maximizing performance while minimizing each customer`s overall system cost."
In developing the DisplayCheck system, much of the engineering effort was spent on algorithm development. The Image Builder prototyping environment from National Instruments enabled ETC`s engineers to quickly fine-tune and experiment with new algorithms. "This is by far the most advanced and easiest to use image-processing `sand-box` application I have ever used," says Pietrzak. Once finalized, ETC used the National Instruments LabVIEW 5.0 and the IMAQ Vision image-processing toolkit to implement the system software.
"The IMAQ Vision image-processing functions are the basic building blocks of the software," says Wayne Lootsma, ETC manager. "These were combined with LabVIEW`s extensive math functions and ETC`s user-interface libraries to build a fully functional and flexible display test program." In addition, the basic test functions of DisplayCheck can be incorporated into even more flexible test systems using TestStand, the latest test executive from National Instruments. Code developed in LabVIEW, LabWindows/CVI, Visual C++, Visual Basic, and other languages can be easily integrated into TestStand scripts to perform multithreaded system-level testing using a combination of motion-control, data-acquisition, and image-acquisition hardware.
To test each display, the vision system determines whether any pixels, rows, or columns are stuck at `On` or `Off,` and recognizes specific patterns or characters. User-defined thresholds allow manufacturers to establish repeatable acceptance criteria for pixel contrast ratio. An important software feature is its capability to locate each pixel in the display. This avoids having to precisely align the display to the camera. To properly locate these pixels, two images are digitized: an image of the display with all pixels turned off and a checkerboard image with every other pixel turned on.
Using an IMAQ routine, the background normalization first compensates for lighting nonuniformity in the checkerboard image by dividing by the `all-pixels-off` image. The checkerboard image is then binarized using a morphology-based adaptive thresholding algorithm, and the pixels are highlighted using blob analysis and filtering routines.
"After this procedure, the pixels in the original image appear much cleaner," says Lootsma, "and their centroids can be determined more easily." Another important system function involves determining whether any stuck pixels are present. After pixel location, an automatic thresholding function determines what the intensity level is for both `on` and `off` pixels. With this accomplished, pixel values are turned on and off and the system can then determine which pixels are stuck on or off. After completing these procedures, pattern-recognition and character-recognition routines can better assess the quality of the display.
While ETC is initially targeting the DisplayCheck system at display manufacturers, subassembly houses, and developers of mobile telecommunications products, the system is not limited to testing displays. "Because we are performing a series of tests on small pixels," says Wilson, "the system could easily be adapted, for example, to test the quality of ink-jet and laser printers."