- Trends in Camera Design Increase Functionality
Higher resolution, more features and faster speeds are enabling novel camera technology while reducing imaging application costs.
Michael McKay, PixeLink
Commercially available camera sensors are shrinking, perhaps driven by the camera-phone craze. Now common are 3-μm pixels with megapixel sensors that measure less than half an inch. Although beneficial for compact cameras, such small sensors are not ideal for machine vision or microscopy. The smaller pixel size reduces sensitivity, increases noise and requires more expensive optics. The benefit is the cost: The economies of scale in the consumer market make these sensors less expensive than those dedicated to scientific or industrial applications.
Sensors also are becoming faster. Whereas 12-MHz pixel clocks used to be commonplace, current sensors with 40-MHz pixel clocks can deliver 1280 × 1024-pixel images at nearly 30 fps. Sensors that can support faster clocks are available, but a more common approach to increasing speed is the use of multiple taps with 40-MHz clocks.
Supporting the new sensors has necessitated changes in the electronic design of cameras. CMOS sensors, in particular, require pixel calibration and correction to provide accurate and consistent response. The corrections can vary from sensor to sensor, so the best place to perform them is onboard the camera. This involves specialized processing hardware, such as digital signal processors, microcontrollers and field-programmable gate arrays. Learning how to best perform the corrections and calibrations has been one of the major challenges for CMOS camera manufacturers, but advances in this area have enabled image quality comparable to that of high-end CCD cameras.
Along with image processing and calibration, improved electronics have enabled more camera functionality and the use of standard digital interfaces. Many of the features one would expect to see on a frame grabber are commonplace on modern cameras. These include onboard memory, look-up tables, image processing filters, advanced and flexible triggering, and communications ports. Furthermore, the standard interfaces, such as FireWire and Gigabit Ethernet, drive down costs by reducing the need for frame grabbers and expensive cables.
Shorten head please
VGA (video graphics array) cameras still dominate in machine vision, but interest in higher resolution is growing. To meet this demand, most camera manufacturers provide industrial cameras with one-, two- or four-megapixel resolution, with a few in the six- to 14-megapixel range.
Although high resolution allows small details to be resolved in a wide field of view, the combination of high-resolution and random pixel addressing available with CMOS sensors is opening new possibilities. Together, these features can reduce the number of pixels processed. Only the areas of interest in a wide field of view need to be imaged and transmitted to the host.
The benefit is that motorized stages and positioning equipment that are used to scan or “step and repeat” with a low-resolution camera can be replaced by a fixed camera with a wide field of view and high resolution. Rather than capture one very large image of a whole circuit board, the camera captures images of only the components being inspected. With the reduced bandwidth, the throughput can be higher than if a full-resolution image is captured and parsed on the host. Imagine being able to replace a robot and control system with a single fixed camera. For OEMs and system integrators, this represents huge savings in system complexity and equipment costs.
The extended dynamic range of CMOS sensors is enabling automation and inspection in areas such as welding and traffic monitoring, where the scene is dominated by high contrast and varying illumination. By creat?ing nonlinear response at the sensor, detail in very bright areas as well as in dark areas in the image can be detected. Depending on the sensor, the nonlinear response can be defined as a piecewise linear curve where the “knee-points” in the curve and slopes of the piecewise linear portions are defined by exposure times. The resulting image has detail in a wide range of exposure.
This feature enables imaging applications where lighting conditions are extreme, with high contrast and a wide intrascene dynamic range. In a welding application, for instance, this capability could be used to image the weld formation, the cooling weld structure and the joint area ahead of the weld within one image. One camera could track the weld joint, monitor the welding process and assess the quality of the weld. For traffic monitoring, where the varying conditions of sunlight, reflections and shadow create high contrast in the scene that can confuse autoexposure systems, extended dynamic range can ensure reliable image capture over a much broader range of light conditions.
Another benefit of the new technology is CMOS antiblooming characteristics. This is particularly attractive for laser profiling, electronics inspection or display technology inspection, where bright direct illumination or specular reflections are an inherent part of the image. For CCD technology, blooming characteristics can leave nasty streaks in the image, and this type of illumination can be a major challenge.
In a demonstration of CMOS capabilities, an electronic part was placed under an LED ringlight, with the light shining in the camera lens. At left is the scene inambient office light. In the center, the LED ringlight is turned on. Note the antiblooming indicated by the lack of vertical stripes from the bright LEDs. At right,extended dynamic range enables detection ofthe original part and the shapes of the various ranges of the LEDs.
In difficult lighting situations, an LED ringlight is positioned over an electronic component but with the light pointed at the camera (see figure). With the light off, a normal high-contrast image is captured. With it on, overexposure washes out the details in the LEDs, but the electronic part is still visible. In a CCD camera, blooming would make detection of the part difficult. Activating the extended dynamic range allows the camera to image both the part and the details in the LEDs.
Color cameras also are becoming more popular for machine vision applications. Artifacts due to Bayer color mosaics used on single-chip color cameras are less noticeable with higher resolutions and onboard processing. Operators much prefer working with color images, and software can process the color information as effectively as, if not better than, the monochrome intensity alone.
Cameras in microscopy
The trend toward higher resolution has been more evident in microscopy, where the desire for more pixels is often well beyond the resolving capabilities of the microscope objective. This has fostered the development of pixel-shifting technology, which has been routinely used to provide multimegapixel microscopy images. In this technology, electromechanical actuators shift the sensor or optics to capture multiple images of the same scene. Software combines the data to create very high resolution images. In the past, it was cheaper to use this approach than true high-resolution sensors, but today’s selection of cost-effective high-resolution sensors is making this technology obsolete.
Perhaps the biggest advantage of the resolution and speed is the availability of high-quality previews from the camera on a computer monitor. A full field of view preview of a multimegapixel image can be provided at 640 × 480 or 1280 × 1024 resolution at fast frame rates. This allows rapid sample positioning and focusing prior to image capture, and promotes adoption of digital imaging in high-productivity activities such as pathology, where hundreds of slides are examined in a session.
Combining the characteristics of a solid microscopy camera with a machine vision trigger and strobes is beneficial in applications for automated slide scanning. Here, stitching together multiple smaller images creates high-resolution images of an entire slide that provides the context of the sample and possibly even the slide label.
New camera technology is not enabling new applications as much as it is reducing overall system cost by cutting the cost of the cameras, removing the need for frame grabbers and reducing system complexity. Although cost reduction may not be glamorous, it has been and will continue to be the driving force behind changes in camera technology.
Meet the author
Michael McKay is product manager at PixeLink in Ottawa; e-mail: firstname.lastname@example.org.
MORE FROM PHOTONICS MEDIA