Once an image has been collected, it must be passed to a suitable computer which can run the required software routines to determine the product or part status. The way that this process occurs is dependent on many factors but usually it comes down to how fast the line runs in terms of the number of parts that must be handled in any given space of time. Some of the options include:
Smart cameras have a small processor on board which can run image analysis software with a limited tool set.
These systems are very useful for simple applications which do not require complex inspection analysis routines.
The camera imaging platform is also usually quite limited in ithe max. image size which they can process efficiently although speed of image transfer is high which makes them useful on quite fast production lines.
Typical applications which these platforms can handle include:
There are a number of "off the shelf" image processing software packages which use a graphical user interface which makes them relatively easy to deploy for simple to moderately complex jobs.
Some of these packages are supplied free with some camera brands although usually the more complex tool sets are sold separatley at additional cost.
More complex jobs usually require the assisitance of a machine vision partner such as Indyn.
These systems run on an external PC. Typical applications include:
Deep learning systems are a recent innovation. Usually the data sets required to make the system perform reliably will require extensive project management and careful planning.
These systems work well on natural products where normal machine vision software has some difficulty.
These systems also run on external PC and data transfer rates are a prime consideration for effective delopyment.
Indyn has considerable expertise in deep learning as applied to machine vision applications.