Looking forward by looking back
With OEM suppliers introducing specialized smart cameras, frame grabbers, and image processors, developers are finding lower-cost machine-vision and image-processing system integration easier to configure from off-the shelf components. Still, developers must add software expertise to build systems for end users. To ease this task, OEM suppliers this year will introduce a variety of image computers with image-capture, analysis, and display functions bundled with image-processing software-developm
Looking forward by looking back
Editor at Large
With OEM suppliers introducing specialized smart cameras, frame grabbers, and image processors, developers are finding lower-cost machine-vision and image-processing system integration easier to configure from off-the shelf components. Still, developers must add software expertise to build systems for end users. To ease this task, OEM suppliers this year will introduce a variety of image computers with image-capture, analysis, and display functions bundled with image-processing software-development tools. These systems, of course, will cost more than the sum of their parts so that OEM suppliers can achieve profitability.
For developers, bundling functions seems like a solid design approach. Why not purchase a fully configured imaging system and add value by developing software? However, the history of the image-processing industry is replete with failed companies that supplied such systems. For example, some years ago, Recognition Concepts (Incline Village, NV) manufactured the Trapix real-time image-processing system. That product, developed with the latest 64-kbyte RAM, could be configured with up to 4 Mbytes of image memory in a compact 10.5-in. chassis. Another past product, the Mars 232 array processor from Numerix (Newton, MA), could perform a 1-k complex FFT in just 1 ms (fast for 1983).
Many similar, but now merged or defunct, companies produced complete image-processing or array-processing systems. They included Vicom (San Jose, CA), CDA (Waltham, MA), Adage (Billerica, MA), International Imaging Systems (Milpitas, CA), and Gould DeAnza (San Jose, CA).
Back then, such companies were the image-processing business. Few off-the-shelf frame-grabber, image-processing software, or camera vendors existed. Interestingly, companies that manufactured vision products for open systems, such as the Intel Multibus and the DEC Q-bus--Matrox (Dorval, Quebec, Canada), Datacube (Danvers, MA), and Imaging Technology (Bedford, MA)--still exist today. The reason for their survival and the emergence of more than 50 PCI-based frame-grabber vendors, more than 30 camera suppliers, and an assortment of PC-based software packages is that to compete effectively vision manufacturers must adapt to changing technologies and offer novel products that fill specific market niches. Systems companies, stuck with proprietary buses, image architectures, and software, found such changes to difficult to accommodate.
Today, few companies offer complete systems. Of those that do, the products no longer use proprietary and complex bus, software, and image-processing architectures. To remain competitive, hardware and software vendors have adopted well-established bus interfaces, such as the PCI or VME bus. Board-level products developed around these buses can also be easily interfaced to nonstandard high-speed analog and digital cameras. Better still, software development tools, such as Microsoft`s Visual C++, now allow developers to offer a range of easy-to-configure software using ActiveX controls.
With the introduction of image computers, OEM vendors must be careful not to repeat the industry mistakes of the past. Certainly it is highly unlikely that such systems will incorporate proprietary buses. However, to enter the systems business, vendors must fully integrate a variety of OEM components. And, they will have to support their installed systems over a longer life cycle than specified for OEM components. By offering such systems, developers can expect to lower the cost of end-user applications. However, this may limit them to products that cannot easily be upgraded during the system-development process.