The number of different imaging lenses and camera sensors available for machine vision applications has increased a lot in recent years. To get the best from any vision system great care must be taken to select the right camera-lens configuration, the wrong decision can lead to wasted money and less than optimal performance. As pixels continue to reduce in size it is a common misconception that smaller pixels lead to better performance when in reality the situation is far less clear cut. Increasing the resolution is only beneficial if a reasonable level of contrast can be maintained and while theoretical models and mathematical limits are very useful they can be misleading.
The best solution will always depend on the application itself and for that reason real world testing is essential to understanding the true performance of any lens-camera system. Limitations such as manufacturing tolerances and sensor response mean that there is always some departure from theory.