Video digitization inside a camera greatly reduces the chance of electromagnetic and radio-frequency interference.
Jeff Baran
In an industrial fly-cutting application, engineers use an ultrahigh-resolution linescan camera to digitally image the newly exposed surface of a complex part after each machining pass (left). A color gradient shows deviations between the CAD model and the finished part.
Digital cameras are displacing their analog counterparts in many industrial imaging applications-but not all. To understand when they are the best tools for the job, it helps to examine how they work.
Digital cameras convert the analog video signal from a charge-coupled-device (CCD) sensor into digital data inside the camera head. This process occurs prior to clocking the video data out to the host processing system. One benefit is that video digitization inside the camera greatly reduces the chance of electromagnetic and radio-frequency interference (EMI/RFI) contaminating the video signal before digitization, allowing for maximum signal fidelity. Digital video also can be stored in a memory buffer inside the camera.
Once the video is captured and stored, real-time image-processing capabilities, such as on-the-fly auto gain and offset correction and simple image-processing routines, can improve the video before the image data reach the host processing system. By comparison, only rudimentary video correction can be done in the analog domain.
Digital video data are also in a natural format for real-time image processing by a digital-signal processor, personal computer, or other custom electronic processing hardware. The additional processing electronics, though, increase the camera body size, require additional power, and raise the cost of the design and manufacturing processes when compared with analog cameras. Still, a digital data stream can be transmitted without data loss or mutation via the common RS-422 and RS-644 protocols, fiberoptics, IEEE 1394 Fire Wire, and other digital transmission protocols. With a few exceptions, such as fiberoptic data transmission, digital cameras usually require short cable distances, say 3-5 m, while an analog video signal can easily be transmitted across 50 m.
In essence, although digital cameras lend themselves to many applications due to their convenient data format and processing features, analog cameras remain desirable when small camera size and low cost are required.
Matrix versus linescan
Engineers specifying cameras for industrial applications also need to understand the difference between matrix and line scan cameras. Matrix cameras output an entire frame of data, one line at a time. Typical frame rates are 30-1000 frames per second, depending on the horizontal and vertical pixel resolution of the camera. Typical resolutions available are 512 ? 512, 1024 ? 1024, and 2048 ? 2048 pixels. Matrix cameras thus have an imaging advantage in stop-action situations, motion-analysis applications, and object tracking.
Traditionally, matrix cameras have captured and transferred image data with a frame-transfer technique that requires a high-speed shutter to prevent ghosting or blurring when imaging moving objects. Frame-transfer cameras without a high-speed shutter are typical in applications such as high-resolution microscopy, scientific research and development, and low-light-level medical imaging.
Progressive-scan technology solves the shutter problem by acquiring an entire image with a single electronic shutter event within the CCD sensor of the camera. The data are then read out of the sensor sequentially, instead of with the odd and even frame-transfer interlaced formats. Common applications for digital progressive-scan matrix cameras include airborne imaging, security, on-line inspection, machine vision, license-plate recognition, and traffic-control equipment.
Linescan cameras are different in that they scan scenes one line of pixels at a time. To construct a two-dimensional image, they scan objects in motion by imaging from a stationary location while the object of interest passes by. This convenient situation occurs frequently in many applications: a factory where a product is moving on a conveyor belt underneath a camera, a camera mounted on an aircraft flying over land, or fluid flowing past a fixed camera.
These cameras also often have higher data rates than matrix cameras. Hence, they can inspect objects moving very fast or processes that require high product speed for reduced production cost. One Reticon (Sunnyvale, CA) linescan camera, for example, can scan more than 100,000 lines per second with 10 bits of data per pixel.
Another advantage of such systems is higher overall image resolution than matrix cameras. Typical linescan-camera resolutions are either 1024, 2048, or 4096 pixels horizontally by one pixel vertically. As a general rule, this resolution is two to four times that of the matrix cameras, and digital image data are provided at higher data rates. Digital linescan-camera applications include document scanning, check sorting, industrial imaging, web inspection, and semiconductor-wafer and mask inspection.
Inspection goes digital
For these reasons, digital cameras have found their way into a wide variety of applications. Users have a diverse range of equipment options that can vary significantly in technical specifications. Examples include a high-resolution 8192 ? 1-pixel linescan camera, a 1024 ? 1024-pixel progressive-scan matrix camera, or a 4 million-pixel frame-transfer smart camera. Each has an advantage depending on its intended application.
Capture Geometry Inside (CGI; Minneapolis, MN), for instance, uses Reticon linescan cameras in its CSS1000 inspection system-equipment that reportedly reduces costs and processing time when inspecting first-article machined parts with complex internal geometry (see photo on p. 139). The system can reverse-engineer parts that fit within an envelope of 12 ? 10. 5 ? 8 in.
To accomplish this, the machined parts are placed on a setup frame within a standard mold, covered in a vacuum system, and filled with a molding material called Encase-it. The molded part is then mounted on an aluminum base and secured to the worktable. The system operator selects the mold dimensions, color, and part material. The control for an industrial fly-cutting machine then automatically selects spindle speed, cutting depth, and feed rate based on these setup parameters.
In the next processing step, the fly-cutter precision-machines ultrathin (0.0005-0.010 in.) layers from the part. The hardened Encase-it material reinforces the part and provides high contrast between it and the mold material, defining the contours on each layer.
After each layer is machined away, the high-resolution linescan camera digitally images the newly exposed surface of the part. Operating in a multiplexed mode, the camera takes data from two separate analog video outputs from the CCD sensor (odd and even channels) and converts each signal to a 10-bit digital number (between 0 and 1024, depending on the brightness of each pixel). After each pixel is digitized, the data travel to an internal microprocessor that matches gain and offset between the channels.
Data are then reformatted into a single stream of raster order video and clocked out of the camera at a 40-MHz pixel rate. The 8192-pixel resolution of the linescan camera allows scanning the slice surface at 1000 dpi.
The image data then go to the first of two software stages, which accepts and files each layer of image data to ensure that no information is lost. After this, the system inputs the three-dimensional bitmap into the postprocessing stage, which creates points at the intersections of the material and the part on each scanned z-axis plane. The resulting measurement accuracy of the CSS1000 is within 0.0008 in.
Replacing film
In another application-inspecting the shock wave of a projectile fired at supersonic speeds-high-resolution digital matrix cameras replace 12 film-based photographic cameras. Prior to implementation of the digital devices, the cameras took photographs as the projectile passed within the field of view of each camera. The pictures were then gathered, processed, scanned, and analyzed. Digital cameras, though, provide instant image data ready for analysis, as well as greater sensitivity for increased dynamic range.
The Reticon replacement cameras contain a back plane with an analog-to-digital digitization board, an internal high-speed data processor, an image memory buffer, and an Ethernet network card. Using high-speed shutters, the 12 networked digital cameras image the shock wave of the projectile in flight. Each unit has its own address, which gives the host computer system the ability to access captured images at any time across the network.
The equipment then transfers the high-resolution images electronically to a central computer system for processing to determine the aerodynamic properties of the projectile. A convenient Web-browser interface allows the operator to download images in a specified format.
Exploring the future
In addition to increased resolution, digital-camera applications are requiring higher speeds and dynamic range. For instance, these cameras are inspecting sheet aluminum for defects while the product is moving at speeds of 3000 in./s. Document-scanning applications require linescan rates up to 80,000 lines per second at resolutions of 1728 pixels per line. Medical fluid-analysis applications need low-light-imaging, cooled digital cameras that output megapixel resolutions with a dynamic range of 16 bits per pixel (65536 gray levels) and user-selectable clocking schemes.
To address the demanding requirements of web scanning, document scanning, and medical applications, digital-camera manufacturers are improving the speed, feature sets, and price/performance ratio of the equipment. And as transceiver prices continue to drop, Fibre Channel and gigabit Ethernet digital-video-transmission schemes will allow 4-km cabling distances. These signals will be completely immune to EMI/RFI contamination.
Multiple back-plane computer systems will process video from multiple-camera inspection systems. Processing inside the digital camera head will continue to increase in complexity. Future cameras could have their own bus, where custom image-processing modules would be plugged into an open slot. Instead of clocking out digital video data, the camera would output decisions or measurements that have been made based upon the preprocessed video.
Networks of digital matrix and line-scan cameras will continue growing in popularity due to intuitive interface developments and faster data-transmission speeds. Data output rates will likely increase to 160 MHz this year, with a dynamic range of 70 dB.
Perhaps the high throughput of the advanced graphics port slot, now standard on new Pentium-based computer systems, will prove helpful in solving the data-transmission bottleneck.
Nevertheless, digital cameras will continue to dominate in areas of speed, real-time video processing, and precision imaging. o
JEFF BARAN is an applications engineering manager at EG&G Optoelectronics, 345 Potrero Ave., Sunnyvale, CA 94086; [email protected].
Laser Focus World n www.optoelectronics-world.com n April 1999 139