Cost Drivers and other considerations for a successful machine or embedded vision project

Cameras typically take up the lion share of revenue resources in a machine vision system; however, today’s cameras are faster, smaller, lighter, smarter and less expensive. A good quality GigE monochrome camera should cost an end user around 1K, costs will go up as megapixel size increases. 

genie-nano Euro block IO cable straight

When choosing a camera interface keep these points in mind: 

  • GigE Vision: Used in low & mid-end vision systems with less critical speed and timing demands; cost effective solution when full speed is not required. 
  • Camera Link: Industry default choice for higher speed connectivity where limited cable length and high cable costs are acceptable, frame grabber is required 
  • CoaXPress: Newer technology, for applications that require higher speeds, longer cable lengths, a frame grabber is required. 
  • Camera Link HS: Originally designed to overcome the speed limitations of Camera Link for line scan cameras, frame grabber will be required 
  • GigE Vision over 10 GigE:  Built on GigE Vision, faster physical layer and better timing accuracy with much higher power consumption; requires server grade equipment for implementation. 


Depending on the camera and the method in which you wish to obtain or retain the data from the image, a Frame Grabber may be required (see camera interface information above).  Frame Grabbers come in a variety of configurations; choices include; bandwidth and DDR-RAM size, on-board camera controls, the number of cameras that each board controls and the newest innovation adds FPGA programing on the board. A basic frame grabber will cost around $500.00, a frame grabber with FPGA programming capability will run around $2,500 or more depending on the configuration of the board, but keep in mind, the board with FPGA will allow you to reduce your CPU/GPU requirements as your images will now be preprocessed. Using an FPGA enabled frame grabber will likely reduce costs for computer related hardware as your bandwidth and RAM requirements will be much less than with a traditional Frame Grabber.   

Depending on your application, lighting can be a strong contender for vision system dollars.  Good lighting is essential to the success of a machine vision system.  Most vision applications will benefit from line lights or back lights.  Getting the lighting right will save you development dollars.   A good quality 5” line light or backlight can run $450-$850 depending intensity controls and size.  Ask your machine vision expert of integrator about how best to “highlight” your intended object. 

Lenses typically represent the lowest percent of dollars spent on a machine vision system.  A standard machine vision 16mm lens typically runs $100-$200, but a high quality specialty lens, such as motorized or Near IR corrected can run well over $1,000.00.  Be sure to invest in a good quality lens. 

Software can be free when the SDK (software development Kit) from the camera manufacturer is included in the camera purchase or can run thousands when complex licensing schema or specialty vision software is required.  There are many machine vision related software solutions on the market, make sure your machine vision professional explains what the free SDK can do for you before investing in costly software. 

Cables represent the least costly part of a vision system but be sure to get certified cables from a reputable cable manufacturer and ensure your integrator is not cutting corners by using cables made by non-certified manufacturers.  A bad cable can make a great vision system worthless. 

Imaging is now touching every industry, entertainment, marketing, auto, aerospace, packaging etc.  Just think of all those robots being developed!  The newest technology is the “Smart” camera.  These cameras have embedded FPGA software right on the board and can be addressed using flow chart style programming.  No longer will vision engineers need an FPGA programmer to complete the system, with a little training a vision engineer will be able to program camera controls, enhance images and much more, all from their desk top or mobile device.  It is truly an exciting time to be involved in machine and embedded vision. 

If you need more information or assistance with a current of future project please contact us at [email protected].  And please, visit our website we are working hard to make it informative and easy to use.  







Imaging Problem? A Band Pass filter might be your answer.

MidOpt_Choosing_a_Filter-Imaging Problem? A Band Pass filter might be your answer. To find the right filter for the job, a broad spectrum white light and a bandpass filter kit will help you highlight various wavelengths. During testing each bandpass filter will achieve similar results as the matching LED wavelength. This process will help you to determine the appropriate LED wavelength necessary for your vision application. By viewing each resulting image each banpass provides, you can determine the best color for maximizing contrast and reduce interfering light. Your vision system may also benefit from other filters to help reduce glare, remove saturation or balance color.

For more in depth information on filter use for your vision system, visit this article link from Mid West Optical, More Info

If you need additional assistance or you need to add a filter swatch kit or specific filters to your image lab please reach out to us at Contact Us


Understanding Back Illuminated Technology in machine vision

Until Hideo Yamanaka published his 2009  “Method and apparatus for producing ultra-thin semiconductor chip and method and apparatus for producing ultra-thin back-illuminated solid state image pick up device”.  The technology of back illuminating sensors was costly, complex and required further refinement to become a widely used sensor.  Yamanaka found that by rearrangement of the imaging elements, increased light could be captured and thereby improve low-light performance.  However, back thinning lead to a host of other problems such as cross-talk which causes noise, dark current and color missing between adjacent pixels.  Thinning also made the silicon wafer more fragile.

A traditional front illuminated camera is constructed to mimic the human eye; a lens at the front and photodetectors at the back, a back illuminated sensor arranges the wiring behind the photodiode substrate layer by flipping the silicon wafer and then thinning its reverse side so that light can strike the photodiode layer without passing through the wiring layer.

Front vs Back Illuminated CMOS

Front illuminated CMOS sensors vs. Back Illuminated CMOS.

Today back illumination technology has made some significant progress and BI chips are now available from several manufactures of Silicon Chip technology.  With higher sensitivity over a broader spectral region (deep UV to near IR) several industrial camera manufacturers are introducing back illuminated cameras.

These cameras are ideal for ultra-low light applications like Astronomy, Spectroscopy and biological imaging.  With back illumination, low light applications get increased Quantum Efficiency up to 95% and lower read noise at <2 e-rms.  When every photon counts a camera with a back illuminated sensor should be your one and only choice.

If interested in incorporating a back illuminated camera into your project check out the new .pco Panda 4.2 BI :  PCO. Panda 4.2 BI






5 Reasons You Should Choose a 10 GigE Camera for your Machine Vision Application.

There are many types of Ethernet cameras available for machine vision applications. This is a brief introduction to Ethernet and why 10GigE cameras are attractive. Continue reading “5 Reasons You Should Choose a 10 GigE Camera for your Machine Vision Application.”


What is the Best Machine Vision Camera for my Application?

How to select the right camera for your machine vision application?  I would like to present a brief description of the basic procedures that my company, Pyramid Imaging, uses when asked by our customers to determine the best camera for their application.  We’ve been doing this type of work for decades and have developed a simplified but reliable way to narrow down the broad array of possible cameras to a small group of possible candidates.

But, let’s talk about some basics.  We consider a machine vision camera as a camera that will be used to obtain images for an automated process.  We need a camera that has the right number of features that will allow us to control it properly and obtain the images that we require.

First and foremost is to understand the application’s goals.  Are we looking at an Inspection process looking at labels and barcodes, are we requiring Metrology to be done; that is, making accurate measurements on an item?  Maybe we are examining high speed events or do we just need “pretty pictures” for presentation purposes?

Application Goals

Once we truly understand the goal for using a machine vision camera we need to now list all of the application constraints.   We are now looking for the sweet spot in the constraint analysis in which the best camera will satisfy all the requirements.

Application constraints

There are literally thousands of cameras to down select to the one that will be best for your machine vision application.  The broad categories for these cameras include cameras that are:

  • Line Scan – 16K Linear array
  • Area Array – 80 MP and higher
  • High Speed – hundreds of thousands of frames per second
  • Analog (RS-170), FireWire, USB 3, GigE, Camera Link (CL) , Camera Link HS, CoaXpress

These cameras area all about bandwidth and cable length.  Remember that high resolution and high speed cameras need high bandwidth requirements!

You also need to know what lens mount the camera should possess.  These include:

M12, CS, C, F, M42, M75 Lens Mounts.

These are listed in order of increasing aperture. Remember that the bigger the sensor the larger the camera opening needs to be for a lens with a larger aperture.  Also, note that for a line scan camera using a linear array sensor the sensor’s length would be used to determine the largest image circle needed from a lens.  For an area array camera you would use the diagonal length of the imager.

There are also various features that might be required in a camera.  Features such as:

  • Selectable Regions of interest(ROI)
  • FPGAs
  • On camera memory
  • Many others

Here is a type of decision tree that can be considered when trying to select the best machine vision camera:

decision tree for camera

There are many factors to consider when trying to select the appropriate and best machine vision camera for your application.  First and foremost you need to analyze and reanalyze the goals for the machine vision application.  The next vital consideration is to use the best lighting and illumination to highlight the objects of interest.  You want to create as much contrast as possible between what you  want to see and try to make everything else disappear.

So, a quick set of steps for selecting the right camera would be to:

  1. Calculate Resolution- smallest detail in the field of view.
  2. Calculate the object or camera speed- determine best exposure time and frame rate to “freeze” the image.
  3. Encoder or speed sensor ?  If yes prefer line scan !
  4. Distance from computer or display ?  Resolution, Frame rate, Distance dictates data bandwidth , video protocol and cables.
  5. Time to record ?  Dictates amount of RAM and/or RAID storage.
  6. SWAP and other constraints ?  Prioritize
  7. Narrow down and sort camera list on priorities- features, cost, lead time, technical support, viability of manufacturer, etc.

Pyramid Imaging has been providing assistance to customers for decades on selecting the right cameras and other machine vision components.  Go here to see a small example of some of the projects in which we’ve been involved.

Pyramid Imaging  provides this very convenient tool that you can use to  down select cameras based upon certain specifications.  Just go to the camera selector and click on the specifications desired.

We’d be happy to provide you with our free assistance should you like to discuss your application with our experts.  Just Contact Us.