How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
Affordable Flat Panel HDTV: Delivered to your living room with vision
As the demand for flat panel displays increases, the need for faster, more efficient inspection solutions is growing as well. Is the machine vision field ready to meet the challenge?
by Philip Colet
Advances in illumination, optics, sensors, vision processors and vision software are giving semiconductor equipment manufacturers the components required to solve today's machine vision challenges. One of the most crucial of these challenges is the need to generate and process a much greater volume of data at higher speeds. This challenge has arisen because of the need to detect smaller FPD features and potential defects, which requires higher resolution sensors and shorter wavelength lighting. Additionally, each high-resolution image is being gathered over a larger surface area, thereby generating greater quantities of data that must be acquired and processed more quickly in order to keep pace with high-speed manufacturing lines.
Recent advances in lighting and optical technology are available for generating-and collecting-short wavelength (ultraviolet, or UV) light. Sensors from companies like DALSA and Kodak Imaging are approaching UV sensitivity, while UV laser illumination devices are also available. As any seasoned machine vision specialist will tell you, a well-designed lighting system will eliminate 90% of the data in an image-which is a good thing! This is because most of the image data is not critical to developing an understanding of the image itself. As an example, look at the two images on page 24. The image on the bottom is easy to read, while that on top might require some image processing in order to derive the "information" contained in the image.
The only difference between the images is the lighting technique utilized for illumination. The image on the bottom utilizes a technique called "dark field illumination," which is commonly used to enhance specific defects while reducing background areas of the image.
Lighting and image collection/sensing is only the first component of any machine vision system. The second component is data analysis, with the goal of finding specific defects that could lead to future FPD failures. Here, the challenges lie in the time available for processing and the quantity of data generated by the imaging system. The latter can easily reach into the Gigabytes of data, gathered over the course of a couple of seconds. Certainly, prudent lighting choices can reduce the amount of effective data tremendously, but many hundreds of Megabytes of data will still need to be processed in a given span of time.