Q+A CENTER

All PC-based imaging systems incorporate common components, most notably a camera. Chances are that you will also need a device for streaming the image data to your computer’s system memory, a role that is fulfilled by video capture cards, more commonly known as frame grabbers. This article discusses the most common interfaces in use today and how they are used in an imaging system.

Analog and Digital… and Digital

Generally, you will select the camera first before anything else. Although this is step one, it is by no means a simple task. Choosing a camera means determining the needs of your application and surveying the vendors. Among the basics, you must determine the required resolution, speed and sensitivity. Will your image be in the millimeter or micrometer order of magnitude? Will the camera take pictures of objects zipping along a conveyor belt or cells on a microscope’s slide? How much light is needed for acquiring an image? If you buy from a well-known camera manufacturer, the camera model you select will likely be available for different interfaces. It is the interface that will determine your image or video acquisition hardware. The most common standard interfaces in use today are analog, LVDS, IEEE 1394, Camera Link™ and, more recently, Gigabit Ethernet.

• Analog. Traditional frame grabbers that convert analog signals to digital still have their place in machine vision applications. They offer stability and support long cable lengths. Entry-level analog frame grabbers feature off-the-shelf video decoder chips for digitizing the color or monochrome video signal in NTSC/RS-170 or PAL/CCIR formats. Non-standard analog formats can be supported with higher-end frame grabbers, but you will need to configure your card to recognize the camera’s video signal.
• LVDS. A digital interface based on an analog signal format, in that image data is transferred in raster form. Transmission and cable lengths are similar to that supported by Camera Link®
• Camera Link®. Camera Link® technology’s high-performance and standard cabling revolutionized machine vision a few years ago. Based on LVDS, Camera Link® is also raster-based and operates in a number of modes. A Base mode video acquisition card features a single deserializer chip and supports a maximum data rate of 2.04 Gbps. In Medium mode, with two chips, the maximum data rate is 4.08 Gbps. Finally, Full mode, with three chips, the maximum data rate is 5.44 Gbps. For all modes the maximum cable length is 10m.
• IEEE 1394 or ‘FireWire’. This network link was developed by Apple and the IEEE in the late 1990s. Now in its second revision, IEEE 1394b has a peak transmission rate of 800 Mbps. Its cable length depends on the medium. Copper wire tolerates a maximum 4.5m cable, but plastic and/or optical fibres can handle longer lengths without data loss. Data transmission is packet-based and bi-directional. Generic adapter cards are used to stream the image data to system memory instead of a video acquisition board.
• Gigabit Ethernet (GbE). Like IEEE 1394b, GbE is a network-link that is ready for a big splash on the imaging/machine vision scene. As its name implies, GbE transfers data at a peak rate of 1Gbps over a maximum distance of 100m. GbE cameras connect to a standard Network Interface Card (NIC) to stream the image data to system memory. The Automated Imaging Association recently released a machine vision protocol standard for Gigabit Ethernet, GigE Vision™, and Matrox Imaging was among the first imaging vendors to support this protocol standard on current products.

Depending on the particulars of your application, you might choose a smart camera. A few short years ago smart cameras descended on the marketplace. While they are not revolutionizing image processing per se, they are a different vehicle for existing algorithm technologies. What smart cameras offer is a new package of delivery; they open opportunities to systems integrators because they are highly-integrated devices with a small footprint. For applications with space restrictions, smart cameras can resolve the issue. Furthermore, devices (smart cameras) with their higher level of integration, might offer lower-cost solutions to the user.

All new image processing technologies can be run on a smart camera, namely geometric pattern recognition and edge-detection, the edge-based image processing techniques that relax optical and illumination requirements and render sub-pixel results. But smart cameras are limited in processing power, and they might not be fast enough to process complex imaging algorithms in a practical way. Due to power consumption issues and heat dissipation, smart cameras cannot be fitted with the same processors as a PC; there is a limit to the types of applications that can be run on a smart camera.

Imaging Hardware’s Capabilities and Caveats

Once your camera and interface are established, you are ready to choose your imaging hardware. This could be a traditional frame grabber, digital video interface card or vision processor. The particulars of your application and of course, your budget, are the deciding factors at this stage.