How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
By Lee J. Nelson
Independent of the application, when acquiring imagery for subsequent analysis, whether by eye or by machine, illumination is fundamental. Apposite lighting imparts context and is basic to understanding image content. If shadowed or obscured, even seemingly familiar objects can confound or be misinterpreted.
Take, for example, automatic license plate identification with its objective of extracting and reading foreground alphanumerics. The extensive and heterogeneous population of North American plate designs is apt to baffle even the most robust video-based recognition engines. Imaged in visible light, background graphics often interfere with the letters and numbers, impeding interpretation of the latter.
Conventional CCD cameras may be blinded by headlight glare. When a detector element becomes saturated, the stored charge “spills” into adjacent sensor cells. The outcome is a smeared picture in which alphanumeric characters either are illegible or indistinguishable. Video cameras with automatic compensation generally handle saturation by increasing shutter speed until the headlights are exposed properly. The gain, however, is so reduced that the license plate effectively disappears. And, if the vehicle moves during integration time, the result is motion blur.
Researchers have experimented with optical bandpass filtering, digital spatial filters, and by varying camera and illuminator positions relative to the target plate as well as to each other. All conferred limited benefit due to ever-changing operating conditions: time-of-day, weather, ambient lighting, headlights, tail lamps, etc. A solitary configuration simply could not address all scenarios.
A potential solution may lie in shifting wavelengths into the near-infrared (λ = ≥750nm) and by providing as nearly as possible a constant level and direction of illumination. Clearly, the camera has to be sensitive in the infrared region of the electro-magnetic spectrum and it must be outfitted with a filter that rejects visible light. Many North American plates are partially retro-reflective: light rays from an infrared illuminator bounce off the plate and return directly to the camera. Without any stray visible light or other reflections or refractions, the resulting image exhibits a gray background and lacks detail except for sharply contrasted alphanumerics. By inverting all pixel intensities, the foreground characters are segmented and prepared to be identified by the recognition engine.