Bookmark and Share

General Discussion

Machine Vision Too Expensive
3 replies [Last post]
Russel's picture
Last seen: 2 years 2 days ago
Level 1: Prestidigitator
Joined: 2012-07-25
Points: 2

Over the years that I have learned of machine vision, and understood its application, I have made many inquiries on their costs but I have unexpectedly found them to be too expensive - and far much out of reach. Yes, they provide such important solutions, but would you, for example, willfully buy a device costing almost half the cost of your vehicle to install in the vehicle?
With me is a set of improvements and features I would like to be added to an ADAS but I am not a software developer. The whole system, if well implemented, will be able to meet the needs of a huge market - in many nations, which are for sure waiting for its arrival, and the additional features I'd like to add can be added to the existing ADAS.
China is well known for availing devices in Africa that are within the reach of the multitude, and thus there is such a large market for them, that's why they've become such strong trade partner with many African nations; even though I haven't seen them putting much interest in machine vision.
Is there an EmbeddedVision member who could be willing to work with me to develop the system, my main interest being to provide a solution to this needy market? I don't intend to develop it as a stand-alone ADAS, but to use it together with other devices.
Thank you.
John Russel Ouko, a fifth year student at Egerton University taking BSc. Instrumentation and Control Engineering.'s picture
Last seen: 1 week 1 day ago
Level 1: Prestidigitator
Joined: 2011-06-14
Points: 2

Another thing to consider when targeting low cost embedded systems is the algorithm complexity, optimization, and performance. Can you target a "boxed" or "generic" algorithm or do you need an algorithm optimized/tailored to a lower cost processor? 

We (Twisthink, LLC - bring value to our customers by collaborating with them throughout the algorithm development process to deliver a highly optimized algorithm that meets the price target and performance target. We get down and dirty with the application and identify key succuss factors, noise sources, and optimal detection techinques. We have developed several cutting edge algorithms that offer first-to-market functionality by leveraging our background in mathematics, statistics, signal processing, and information theory. 

Hardware is one piece of the puzzle, but don't forget to balance this with algorithmic complexity and system performance.


Ryan Johnson

Jeff Bier
Jeff Bier's picture
Last seen: 4 weeks 10 hours ago
EditorLevel 4: Thaumaturgist
Joined: 2011-05-29
Points: 93

The cost of processors and sensors for embedded vision applications is rapidly falling.

We can see this because now automobile manufacturers are integrating vision-based ADAS systems not only into luxury cars, but also into mid-range models, like the Subaru Forester.  I wrote about this in my recent column here:

There is even an ADAS app that run on smartphones, called iOnRoad:

Those interested in designing low-cost vision systems might consider starting with a low-cost vision-oriented development kit, like the FinBoard from Avnet and Analog Devices:


Last seen: 2 years 17 weeks ago
Level 1: Prestidigitator
Joined: 2013-04-18
Points: 2

Industrial machine vision systems tend to be very high performance and are often customized to an application.  By their nature, they are not sold in the very high volumes that you find with products for consumer applications.  This is why they tend to be expensive.  Their expense, however, is very justifiable when you consider the increases in throughput and quality that they allow, as well as the savings in labor costs that they provide.