Automatic selection of image features for visual servoing

The problem of automatically selecting the image features is an important aspect of visual servoing systems. This paper addresses the main issues relevant to automatic features selection in the context of visual servoing. The constraints imposed by the task, the feature extraction and state estimation processes are derived. The quality measures are formulated and a features selection strategy is introduced. In addition to the features selection, the proposed methodology provides the automatic selection of windows for the target features during the servoing. Also it can be applied for optimal vision sensor positioning. The design and implementation of an automatic CAD-based features selector (AFS) using the proposed methodology has been achieved. The experimental results are presented to demonstrate the effectiveness and correctness of the proposed framework.