Part 2: The Vision Chain

In Part 2 of our Guide, Multipix introduce ‘the Vision Chain’ and the components that are required to create a successful Vision Solution

Firstly there are two main approaches for creating a machine vision solution, which one is used will depend on the image processing demands and situational use of the inspection system.

The Vision Chain

Outside of the Vision Chain, but essential to the solution, is also communication with external systems such as PLC’s and this will be via I/O, Ethernet, Serial comms or similar.

Pros = Flexibility    Cons = Complexity

The Smart Vision Chain

Comprises of an all-in-one Camera with Lens, integrated processor, image processing tools, I/O and some also have built-in Lighting.

Pros = Ease of use   Cons = Restricted performance

Let’s take a look at each component in more detail:

Lighting:  The importance of illumination is almost always under-estimated in machine vision applications. Images depend totally on the light reflected and absorbed by an object which are then ‘collected’ by the sensor/camera that creates an image. It makes sense therefore that the lighting technique used has a significant impact on the image captured. Common lighting schemes are explained here

Using CDI – Cloudy Day Dome Light has dramatic effect on this reflective surface!

Lens:  Quality varies dramatically as do the features of a lens. A megapixel camera for example requires a high resolution lens and many applications benefit from the use of telecentric lenses. The wrong choice can result in bad image artefacts such as blurred images, distortion and vignetting.

It’s important to work our your lens size using the following formula.  You can look up your sensor size here


Filter: Used to attenuate parts of the spectrum before they enter the lens and so prevent the camera capturing the light from that part of the spectrum. The visible spectrum is most common in machine vision solutions although Near Infra-red and Ultraviolet also feature

For example, say the solution is using a red ring light (635nm wavelength) for lighting the object to be inspected, it then makes perfect sense to block out all other light from the image by using a 635nm Red Bandpass filter. This ensures other light from other sources such as daylight/overhead fluorescence do not affect the image being captured. Image stability is key to ensuring a stable inspection system!



Camera:  This has a sensor fitted to take an image, usually on demand by an external trigger signal. Specifications of a camera include frame rate, resolution and sensitivity. Features of a camera include gain and exposure control, noise reduction, colour enhancements and so on. There are many different cameras to choose from and this is one area Multipix discuss at length with customers to ensure they are advised on the best option(s).

Some common points to consider when selecting a camera are:

  • Frame ratehow many objects to inspect per second?
  • Resolutionwhat is the smallest defect that must be measured or detected?
  • Is colour required? – most inspection tasks do not involve colour!
  • Interfacehow far is the camera mounted from the processing PC? Multiple cameras?
  • Areascan or linescan camera?

Framegrabber: These are additional cards, typically installed in a PC platform, that are required to connect the camera to the PC for camera control and image transfer. Once the image has been transferred into PC memory it is available for processing. Framegrabbers are designed for efficient image transfer with zero or minimal CPU overhead. It is common to apply a trigger signal to the frame grabber that then triggers the camera to acquire an image. In some cases when using USB3.0 or GigE it is possible to use the native motherboard connection and directly trigger the camera.

PC: The most common platform for a vision solution. They can be in a 19” rack industrial format, traditional desktop, shoebox/mini PC, backplane and so on. Many solutions run on Windows/Embedded Windows but Linux is also an option.


Image Processing SoftwareThere are different levels of user complexity when it comes to the image processing software which is usually related to how sophisticated the image processing task is. Other factors include the skill level of the machine vision developer using the software, the cost of the run-time license and the total quantity of systems being deployed.

There are two main choices:

  • image processing library consisting of 1000’s of functions, where software language skills (C++ , C# ….) and know-how are required to develop a solution;
  • all-in-one software products for quickly building machine vision applications without programming.

Which approach to use is heavily dependent on the circumstances and it is an area that Multipix Imaging spend considerable time and effort with customers to explain the options clearly and offer training where required.

Smart Camera: Having now explored the Vision Chain as individual components, it is easier to understand why Smart Cameras often require significantly less development time to create a vision a solution. The downside is limited flexibly and constrained processing power and that is why both vision chain approaches are equally popular.

Smart Cameras offer machine vision functionalities embedded in stand-alone device which makes them easy to install in manufacturing environments and are normally IP67 rated, so ideal for food/beverage applications.

One such Smart Camera available from Multipix Imaging is the Datalogic P series which product highlights including;

  • Fully integrated ultra-compact device
  • Rotating connectors for 0°& 90° form factor
  • Rugged IP67 rated housing
  • VGA and 1.3 megapixel resolutions
  • Gray-scale and colour CMOS sensor
  • Interchangeable lenses, illuminators and filters
  • Built-in Serial and Ethernet interfaces
  • Powered by IMPACT LITE software

Read more in Part 3 : Understanding Machine Vision Camera Technology.  Which also considers Area Scan V’s Line Scan?  Which is best for your application?

 Part 1: Introduction to Machine Vision Part 3: Machine Vision Camera Technology