Memories are made of this

The revolution in electronics and machine-vision performance needs to be matched by products that prove they are different.

Unlike conventional von Neumann-based computer architectures, the distributed processing capability of the human brain is intricately intertwined with distributed memory and sophisticated I/O mechanisms. With this type of architecture, human beings can process very complex tasks based on long- and short-term memory in what seems like a very short period of time.

While many computer designers have tried to emulate this architecture, the progress of today’s computer systems is also based on attempts to balance processing power, I/O capability, and memory to increase throughput of computationally intensive tasks. With the introduction of dual-core processors, high-speed interconnects such as PCI Express, and fully buffered DIMMs, Intel and others are building on existing architectures to gain order-of-magnitude performance increases of computer-based systems.

Currently, OEM suppliers are working to leverage these developments in their next generation of products. Epix (Buffalo Grove, IL, USA; www.epixinc.com), for example, has demonstrated a singe Dual Base PCI Express Camera Link frame grabber interfaced to two Basler (Ahrensburg, Germany; www.baslerweb.com) A504kc cameras, both running at 500 frames/s. What was impressive was the relatively low cost of the frame grabber and the fact that no on-board frame-grabber memory was required to buffer the captured images. According to company president A. C. Petersen, Epix is also at work on a single PCI Express board that will support up to eight Camera Link cameras using a custom Camera Link-to-frame-grabber board interface.

One of the key beneficiaries of these developments will certainly be the developers of systems that use image-processing and machine-vision equipment. No longer will these companies require multiple frame grabbers to support multiple cameras. However, the impact of advances in processing, I/O, and memory capabilities will not simply affect developers of frame grabbers and those that deploy them. Developers of machine-vision and image-processing software will also leverage the performance of dual and quad core CPUs in a number of ways, either running multiple instances of their software on multiple processors to perform different tasks or partitioning individual processors for single tasks.

For designers of today’s smart cameras that incorporate image sensors, processors, memory, embedded software, and I/O interfaces, such developments present the greatest challenge. The designers must keep abreast of all these developments and cleverly incorporate them into their products in a modular fashion. To date, many smart- camera vendors are doing this by building cameras using interconnected modules that partition the functions of image capture, processing, and I/O.

For many OEM suppliers to the image-processing and machine-vision market, however, these hardware improvements, while important, will not prove the key to a successful sale. More differentiation is needed. Visitors to last month’s VISION 2006 show in Stuttgart, Germany, witnessed an array of cameras, frame grabbers, software, and lighting products. But in what seems like an overcrowded camera market, the differences between specific cameras were not immediately apparent, with many using similar CCDs and digital interfaces. With no standard performance benchmark with which to evaluate them, this situation will continue.

In the future, it will not be enough to tout the faster speed and flexibility of these products. Nor will prepackaged demonstrations of systems operating at extremely fast rates unachievable in real-world situations suffice. To properly market OEM products, vendors-especially those in Asia and North America-will need to emulate their European counterparts with a greater customer-oriented approach, third-party testimonials, and more sophisticated brand-awareness programs. Indeed, as technology drives OEM products to become commodities and consumer-like items, it will be the service providers who will prove to be increasingly important in the machine-vision food chain.

Andy Wilson
Editor
[email protected]