Validating Autofocus Algorithms with Automated Tests

• DOI: 10.3390/robotics7030033
• Publication Date: 2018-06-25T15:03:25Z
• Authors: Tobias Werner; Javier Carrasco
• Keywords: Autofocus; Data set

For an automated camera focus, a fast and reliable algorithm is key to its success. It should work in precisely defined way for as many cases possible. However, there are parameters which have be fine-tuned it exactly intended. Most literature only focuses on the itself tests with simulations or renderings, but not real settings. Trying gather this data by manually placing objects front of feasible, no human can perform one movement repeatedly same way, makes objective comparison impossible. We therefore used small industrial robot set over 250 combinations movement, pattern, zoom-states conduct these tests. The benefit method was objectivity monitoring important thresholds. Our interest laid optimization existing algorithm, showing performance benchmarks This included standard use worst-case scenarios. To validate our method, we gathered from first run, adapted conducted again. second run showed improved performance.