Motivation
Opteon was founded to address the critical need for sources of extremely high fidelity imagery, perfectly coordinated with action in the real world, to fully realize the benefits possible from applying the best machine vision and image analysis techniques to microscopy, biological sciences, agriculture, and factory automation.
Machine Vision is a seductive …
It is easy to see how well many practical real world problems can be solved if computers could receive, analyze, and act on the information available from accurate, timely images of physical systems. We can literally see the answers with our own eyes.
… but ruthlessly unforgiving discipline
What is easy to overlook, until your project runs aground on them, is just how many factors must be controlled, and to what precision, for machine vision results to be accurate and reliable.
Images must be precise. In any analysis, the better the data, the better the answers. This is particularly true of Machine Vision. If the data is faulty, the analysis may be useless, or worse – wrong and misleading. Accurate machine vision requires:
- Perfect control of lighting with sufficient brightness and the correct light field
- Optimum sensor selection, including pixel size, charge well depth, and sensitivity
- Lensing that matches the application to the sensor
- Extremely low noise electronics
- Near zero system-wide timing latency
Image data must be transferred on time and with delivery of every pixel guaranteed.
Computations must be performed quickly enough to keep up with the machinery. For simple inspections a small local ARM processor may be sufficient, but in very demanding cases a remote server incorporating dozens of cores may be required.
Answers and decisions must be resynchronized with the real world to be effective.
It is never sufficient to consider a camera in isolation. It must be assessed as a system element – critical requirements for which are often missing from other cameras.