How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
Automated optical inspection of a wide range of products requires the camera lens to refocus many times to acquire images at several target positions on the object. The speed at which the production line can run is often limited by the speed at which the autofocus mechanism works. One such autofocus-speed-limited application is flat-panel display inspection.
Substrates for flat-panel displays are now typically two meters across and less than a millimeter thick. The wiring patterns they carry, however, are microscopic. Inspecting them, therefore, requires moving a video microscope mounted on a gantry to a large number of inspection points covering the patterned area. With so many fields to inspect on each panel, the time used up at each stop creates an inspection bottleneck.
With today’s high-speed image-processing engines, actual inspection time – the time to acquire an image and locate any defects in the field of view – has become quite short. It is, in fact, much shorter than the time needed to perform other tasks that are part of the image-acquisition process: slewing to the new inspection-target position, acquiring a new image, and processing the new image. It should be possible, therefore, to reduce or eliminate the inspection-time bottleneck by reducing the autofocus time at each target position.
Traditional autofocus systems use multiple images acquired at various focus settings to locate the position of best focus. Being essentially digital in nature, they suffer from two ambiguities: distance and direction. That is, they cannot – with a single image – determine how far the camera needs to move to reach perfect focus, nor can they even determine the direction to move to improve focus.
To resolve these ambiguities, they use a method that is essentially digital in nature. They iteratively make a small focus change, then calculate the spatial bandwidth of the acquired image, then compare this bandwidth with previously acquired images to decide whether to make another change. The presumption is that better focus provides finer details that appear at higher spatial frequencies. So, the feedback system attempts to find the focus position that maximizes the limiting spatial frequency by trial and error.