Andrew Wilson, Editor, [email protected]
With the introduction of large-format CCD and CMOS devices with ever-decreasing pixel sizes, optics manufacturers are continually being challenged to develop lenses to support them. Unfortunately, this is not an easy task and, for developers of high-performance imaging systems, one that may necessitate using more expensive, high-performance optical systems.
Indeed, some of the high-resolution CCD imagers now available are already pushing lens designers to accommodate these smaller-pixel pitches. Sony’s ICX432DQ/DQF for example, is an interline-transfer CCD featuring a 2088 × 1550 CCD imager with 2.8-µm square pixels. Such small-pixel imagers are being used by companies such as PixeLink (Ottawa, ON, Canada; www.pixelink.com), which has developed a 2208 × 3000 color FireWire camera based on a CCD device with a 3.5-µm pixel pitch.
While the parameters of these detectors and cameras are impressive, purchasing a lens to resolve images at such a high resolution is no easy feat. If, for example, the system is designed to resolve 125 line pairs/mm, then the pitch spacing of each individual line will be 4 µm. To resolve this, the Rayleigh criterion dictates that a CCD or CCD camera using a detector with a 4-µm pixel pitch must be used. Similarly, a detector with a 2-µm pixel pitch should theoretically be able to resolve images with 250 line pairs/mm.
To resolve such images, lenses must be coupled to the camera body. In choosing a lens for any particular application, the modulation-transfer function (MTF) of the lens provides rapid insight into its resolving power (see photo). At frequencies at which the MTF of a camera is 100%, the pattern is unattenuated and the image retains full contrast, theoretically being able to perfectly discern the darkest and the brightest spots of an image. As the frequency of the image or resolution decreases, the lens is unable to resolve these differences as well.
“In fact,” says Donald Ehinger, vice president of international sales with Navitar (Rochester, NY, USA; www.navitar.com), “there are no low-cost commercially available lenses that can be used to leverage the performance of cameras such as those developed by PixeLink. What makes the issue even more complex,” he says, “is that many color sensors such as Sony’s use the Bayer pattern to capture color images.” To ensure that the proper wavelength is properly focused on the imaging array, specialty materials such as fluorite optical glass must be used in lens designs. At present, a number of companies, including Navitar, are developing lenses for high-resolution cameras.
High-performance CCD imagers are especially useful in systems that require large amounts of data to be collected fairly rapidly. Devices such as the KAI-11000 from Kodak (Rochester, NY, USA; www.kodak.com), for example, use interline image transfer to capture 4008 × 2678 images at speeds of up to 2.5 frames/s.
“When using such devices for geo-referenced airborne systems,” says Kevin Kearney, CTO of Geospatial Systems (Rochester, NY, USA; www.geospatialsystems.com), “these imagers must be properly mounted to ensure a high tolerance in all six degrees of freedom.” An example is Geospatial System’s KCM-11, a camera that features the KAI-1000 and transfers 2.5 frames/s at 28 Mbyte/s over its Gigabit Ethernet interface. The camera’s Sensor Management Unit also interfaces to an external, third-party Inertial Measurement Unit (IMU), which includes a GPS receiver and a six-axis position sensor. Supported IMUs include those from Leica (Solms, Germany; www.gi.leica-geosystems.com), BEI Systron Donner (Walnut Creek, CA, USA; www.systron.com), Applanix (Toronto, ON, Canada; www.applanix.com), and NovAtel (Calgary, AB, Canada; www.novatel.com). The resulting system records both image and position-orientation data with each frame.
“In the design of the KCM-11,” says Kearney, “a specially built mount was used to house the CCD so that it could be finely adjusted using interferometry-based alignment techniques.” Designed to image wavelengths in the 680-800-nm Landsat Band 6 spectrum, Geospatial Systems built a custom designed filter for the CCD. “By blocking the blue channel of the sensor’s Bayer pattern, the camera can be optimized to operate in the color near-infrared spectrum,” Kearney says. This feature is especially useful to such customers as the Department of Fish and Wildlife (Washington, DC, USA; www.fws.gov) that wish to monitor vegetation and soil moisture.
In building such cameras for photogrammetric systems, special importance must also be paid to the choice of lenses that are used. “While bayonet-mount lenses from companies such as Zeiss (Oberkochen, Germany; www.zeiss.com) and Leica make excellent lenses for photographic applications,” says Kearney, “they have not been developed for ruggedized applications.” In such applications, very large temperature variations can occur, changing the focus of the lens during operation.
While many bayonet-mount lenses are useful in large format imager-based cameras, they do not have the stability of a fixed-mount device. “Such lenses may also incorporate floating elements, adjustable apertures, and focusing elements-features that are not required in geo-referenced airborne systems,” Kearney says. “Because devices such as the KAI-11000 feature 9-µm pixels, cameras mounted in aircraft flying at 10,000 ft roughly image 1 m/pixel. Therefore any 100-µm movement of the CCD imager or lens will equate to a 10-m ground resolution shift in images captured-a figure that is unacceptable in airborne-based systems.
In the design of the 135-mm fixed-focus,f/2.8 fixed-aperture lens for the KCM-11, Geospatial Systems built a custom lens with no moving parts that is bolted directly onto the CCD mount. Before this occurs, however, the reference flange of the lens must be calibrated with interferometric techniques to ensure that it will be properly aligned with the CCD mount.
“Even after both lens and CCD mount are calibrated and bolted together,” says Kearney, “radiometric and geometric calibration of the camera system must be made.” To perform this testing, Geospatial Systems uses a calibration cage developed at the United States Geological Survey’s Optical Science Laboratory (Sioux Falls, SD, USA; http://calval.cr.usgs.gov/crs/calibration.php).
For each camera calibrated, nine or more images of the 3-D calibration cage are captured from different angles. The cage contains approximately 200 targets, and this yields an abundant redundancy of control points for a rigorous least-squares calibration model. Captured points are then measured and the results fed into a PC-based software program that calculates calibration parameters including calibrated focal length, along with radial and de-centering lens distortion coefficients. “These coefficients are then available for correcting any measured image point for any rectification or precision photogrammetric application that applies,” says Kearney.
To extend its camera offerings in the geo-referenced airborne systems market, Geospatial Systems has partnered with Redlake (San Diego, CA, USA; www.redlake.com) to exclusively develop, distribute, and support Redlake’s multispectral three-chip camera line of digital cameras originally developed by Duncan Technology. Under the agreement, GSI will incorporate Redlake’s technology into its airborne product line.