Network centric vision systems

30 November 2008

Recently introduced standards in the machine vision world make it easier to connect high-speed cameras with control systems to provide more accurate timing.

Conveyorised production line with analogue camera vision system
Conveyorised production line with analogue camera vision system

The introduction of the GigE Vision™ (Gigabit-Ethernet for Machine Vision) and GenICam™ (Generic Interface for Cameras) standards in 2006 heralded a new dawn for vision technology.

At the basic level, they provide the opportunity to seamlessly integrate vision systems into a network environment using industry-standard Ethernet components such as cable and switches, as well as allowing images to be transferred over distances up to 100 metres—significantly further than using other data transfer interfaces. However, with the use of other Ethernet products for I/O and triggering, it is also possible to have a ‘network centric’ approach to running a production-line vision system. As these systems have matured, further developments are coming on stream which provide distributed computing across full multicast server solutions.


Many leading camera manufacturers and vision companies were involved in the development of GigE Vision and GenICam, and continue to be involved in their maintenance and further development, ensuring that they address the needs of industry.

The aim of GigE Vision is to define communication between the camera and PC rather than the functional scope of the camera. The key elements of GigE Vision are a UDP/IP-based protocol and a bootstrap register.

GenICam, which is referenced by the standard, is a central software interface for controlling the camera which allows the devices to communicate their functions to generic software using a standardised XML file. This combination of a uniform protocol and an XML definition makes it possible to use software from any manufacturer for such devices which is hardware-independent, making it easy to ‘mix and match’ the image processing hardware used.


Accurate timing control for conveyorised machine vision systems is essential, including component sensing, camera triggering and reject gate control. This ensures that measurements are made at the right time and that reject product is removed from the process.

Figure 1 (above) shows a conveyorised product line using an analogue camera for image capture. All of the processing is done in the PC, which requires a special frame grabber. There is a trigger signal for machine synchronisation on the PC and digital I/O signals such as the result output also occur through the PC. If the conveyor belt has varying speed an encoder can deal with the changes in speed.

The PC communicates with the camera by means of a serial link and the frame grabber controls the camera and can take in encoder signals as well as having I/O. Although such systems have been in use for many years, they require complicated connections between the system and PC using proprietary cables.

With a GigE camera, however, no frame grabber is needed in the PC since the frame grabber functionality is in the camera and no special cable technology is required. Connections are easily implemented over large distances while trigger and I/O signals remain in the system. This provides ideal separation of image capture from the image processing computer by means of a single network cable. With the system network centric, local and Ethernet I/O can be easily added.

Figure 2 (below) shows a possible configuration for the conveyorised product line using GigE Vision. The computer can now be decoupled from critical timing of the production line. The timing sequence is as follows:

Incoming trigger;

Image acquisition;

Network centric vision system for conveyorised production line
Network centric vision system for conveyorised production line

PC processes image and decides on the result;

Results with inspection ID returned;

Controller waits with result in queue; and

Controller fires I/O synchronised with encoder.

This configuration allows multiple products between the trigger and the reject gate.


To facilitate this configuration local real-time sequencing needs to be maintained. The CC320 Ethernet timing unit from Gardasoft Vision features eight inputs and eight outputs and can provide microsecond-accurate event timings. This allows single or multiple cameras to be triggered at different times and the trigger signal width can be used to control the camera exposure timings.

In addition, the delay between the event trigger and reject gate activation can be controlled, with encoder signals ensuring that the reject gate stays open while the product is at the reject position. The timing for gate activation can be resynchronised to the original ‘component present’ sensor to allow for variations in reject gate timings.

For systems where the conveyor belt speeds and timings may vary, one-, two-, or three-wire encoders can be used to determine the exact movement, allowing the belt to be stopped or even moved backwards while preserving the reject timing.

Some applications require multiple measurements from different regions of the item, using multiple cameras. In these cases, there is often different illumination required for the different measurements, and this needs to be triggered and turned on and off as required.

LED lighting controllers with Ethernet connectivity are also available, offering automatic operation with current-rated and voltage-rated lighting. Control functions include power regulation, intensity control, and the timing and triggering functions required for machine vision systems.


An alternative to image processing in a PC is to use a dedicated processor as part of a compact machine vision solution. The ipd VA61 from DALSA (Figure 3) contains a powerful embedded processor that ensures fast inspection times housed in a rugged enclosure which can be readily integrated into factory environments alongside other automation controllers.

The VA61 supports two dedicated GigE camera ports (in addition to one GigE compliant network port), and the use of an Ethernet Hub enables a large number of cameras to be connected. Multiple camera inputs allow the inspection of different views of the same part, or even different parts, simultaneously. With extensive hardware support for part-in-place detection, strobe illumination, camera trigger, and part counting for delayed pass, fail, and rework part sorting, the need for an additional programmable controller is virtually eliminated.

VA61 GigE Vision compact vision system
VA61 GigE Vision compact vision system


Most GigE Vision software implementations support point-to-point configuration where a single camera connects to a single PC. To maximise the flexibility of implementing network centric systems, the use of multicast enables one camera to deliver images to multiple PCs which greatly speeds up the process of delivering images. Here all the workstations receive the image at the same time, and each packet of the image is sent only once. Once all stations acknowledge that they've successfully received the packet, the next is sent, thus avoiding duplicate transmission of the image data.

Common Vision Blox V10, the latest version of the hardware-independent imaging toolkit from Stemmer Imaging, is one of the only software implementations of GigE Vision to support multicast as standard. In addition it features a GigE Vision Server which enables the computer on which it is installed to behave like a complete GigE Vision and GenICam compatible camera, with freely configurable features. The data output by the CVB GigE Vision Server conforms to the GigE Vision and GenICam standards and is therefore compatible with any standards-compliant software interfaces from other providers.

At its most basic level, images can be transferred from a hard disk or from any CVB compatible imaging hardware via a network interface. In addition local pre-processing of image data can be carried out using a GigE Vision Server PC and distributed computing across full multicast server solutions. The freely configurable GenICam features found in CVB GigE Vision Server provide customised remote control of both the server and the acquisition implemented on it. In a current application, over 250 cameras have been controlled and video distributed using this technology.

—Mark Williamson, Firstsight Vision Ltd.,

See also the article "Integrating machine vision with robots"

See also the article "How machine vision helps deliver zero defects for a fuse manufacturer."


Figure 1: Conveyorised production line with analogue camera vision system

Figure 2: Network centric vision system for conveyorised production line

Figure 3: VA61 GigE Vision compact vision system

Contact Details and Archive...

Related Articles...

Print this page | E-mail this page