Small systems fill a vision gap

In the past, when you needed a machine-vision system, you had two choices: Engage a system integrator to build a system for you, or build your own system. The latter choice involved selecting hardware—a camera, a frame grabber, and a PC—and buying software that you hoped would work with the hardware you had assembled. You can still take one of those two paths, but a newer approach—using an all-in-one machine-vision system—offers a third option at a reasonable cost. All-in-one systems include a computer, built-in frame grabber, and software.

Because an all-in-one system comes prepackaged and ready to use from a single vendor, you don't have to configure hardware and software, and then test, test, and test again to ensure everything works in concert. At most, you'll have to add a camera or cameras and connecting cables, although the same vendors often sell these items. If they don't, they can recommend suppliers of compatible cameras and cables. (See, "Choose a camera, too, ") Truly, these systems are plug and play.

Although all-in-one systems may look like boxy stripped-down versions of large PC-based systems, don't sell them short. The "industrial-strength" packaging of all-in-one systems makes them ideal for use on a production line, more so than off-the-shelf office PCs. And these systems include sophisticated image-acquisition hardware, the capability to work with as many as 16 cameras, and robust image-processing software.

And just because all-in-one systems come in small boxes doesn't mean you can't add hardware and software later on. The Matrox 4Sight II, for example, provides an internal PC/104 expansion bus. Several of the all-in-one systems also let you add application software. In general, though, you should add only software that relates to the system's primary vision tasks. It may make sense to add a statistical process control (SPC) program that immediately reports pass/fail inspection results and charts them for operators. But there's no need to burden the vision system's computer with general-purpose applications such as spreadsheets and word processors. Make sure any software you plan to add will easily link to the software provided with the all-in-one system.

Keep in mind, though, that adding "foreign" hardware or software to an all-in-one system may negate the benefits of buying from one vendor. One reason people choose an all-in-one system is to have a single vendor provide service and support. Thus, if you add to an all-in-one system, you take responsibility for system operation and maintenance. Manufacturers who provided you with a working system will only reluctantly help troubleshoot a system that now includes software or hardware purchased from other sources.

Don't underrate the benefits of working with a single vendor. The vendor's application engineers can give you a quick idea of whether the system will meet your needs.

If you decide to purchase a system, plan to spend time over several weeks working with the application engineers as they learn what you plan to inspect, measure, detect, count, and so on. After they know what you want a vision system to do, they can take samples of a product, acquire and process test images using one of their systems, and offer advice about system setup. As a result of such tests, you'll not only learn whether the system can perform as needed, but you'll also gain insight into the type of camera, lens, and lights you'll need to properly inspect the product. Individual camera, computer, and software suppliers usually won't offer this type of detailed help.

All-in-one systems come with built-in software that's guaranteed to work with the provided frame grabber, digital I/O ports, LAN interface, and so on. You don't have to worry about obtaining the latest versions of driver files or about the compatibility of an operating system with application software. The high-level vision software in most all-in-one system makes it easy to develop machine-vision applications using either click-and-drag or menu-style operations. Most applications won't require you to use C-language algorithms for edge detection, shape matching, metrology, or other operations, because these functions come built-in and ready to use.

How you develop application code varies from system to system, but for most, development work takes place right on the target system. That arrangement makes it easy to use the software's point-and-click or drag-and-drop tools to produce applications and quickly test inspection tasks. The RVSI Acuity CiMatrix Visionscape Express, for example, lets you work with the company's AppFactory software on the target system or on a separate PC. By using a separate PC, you can produce and test new applications without tying up a target system. When it's time to test and debug applications, the software can process images you acquired earlier of the products undergoing inspection.

Although all-in-one systems do work right out of the box, new users should still include several days of training in their implementation schedule. System manufacturers offer seminars that help users better understand the hardware and software. You can attend a multiday course at a vendor's facilities, and the vendor's trainers can visit a facility to perform on-site training for engineers and operators. Obviously, the courses concentrate on using the vendor's system, so class lessons and hands-on exercises relate directly to the product you'll use.

Hands-on training is particularly important for software developers. Machine-vision software provides a variety of tools that can solve a specific machine-vision task in more than one way. As with all software tools, there's no single right way, but many wrong ways there are to apply them. So, for new users, the breadth and flexibility of machine-vision software tools can make applying them properly seem like a daunting task. Through lab work in a vendor's course, developers can become familiar with the basic tools that solve specific types of problems. Having an instructor handy can keep neophytes away from tools that aren't appropriate for their application.

Machine-vision software will include tools that link the system to the outside world. Most systems will communicate with other electronic devices that range from programmable logic controllers (PLCs) that sequence production equipment to networks that communicate test results. At a minimum, a vision system should provide digital lines for control and sensing and provide an Ethernet port to communicate with other instruments and computer systems. The suite of software tools should give you complete control over all I/O ports.

Although RS-232 ports no longer find a great deal of use in new applications, at least one such port comes with every all-in-one system. These days, RS-232 ports exist mainly to communicate with legacy systems. Some all-in-one systems let you specify either RS-232 or RS-485 signals, depending on how you plan to use the serial port.

Before you settle on a specific system, peruse the product's data sheet to ensure you can obtain the number and types of I/O connections you need. If a basic system does not provide all the signals you need, it may accept add-in I/O boards or ports that do.

Even within a family of all-in-one systems, I/O capabilities vary. The Coreco Imaging NetSight system, for example, comes in two versions, one that works with analog cameras and one that operates with digital (CameraLink) cameras. The analog version includes optically coupled I/O lines: eight for output, eight for input, and one for camera control. The digital-camera version lacks the optically coupled I/O lines but adds a parallel port. Both versions offer two serial I/O ports and an Ethernet port. You can't add digital I/O lines to the digital-camera version of NetSight, but you could control the parallel port and use its lines for digital I/O.

All-in-one machine-vision systems may not offer all the options of a custom system, but the tradeoffs of flexibility for single-vendor support make them attractive for many users who want to simply solve a problem. And because the all-in-one systems work right after you unpack them, they reduce expensive development time and help ensure the success of machine-vision projects.

Titus, Jon, "Software makes machine vision easier, " Test & Measurement World, October 15, 2001. pp. 17–20.