Active Vision with Self-Evaluation.
Kenichi Kanatani · Journal of the Robotics Society of Japan · 1997
Active vision for constructing a three-dimensional model of the environment from images requires a robot to control its own motion.Since noise exists in images, the information they provide is not always complete.If the motion is too small, images can provide only 2-D information without any depth clues.As the motion increases, we obtain incomplete 3-D information, which we call 2.5-D information.After the motion becomes sufficiently large, we obtain complete 3-D information.We give a geometric interpretation to these transitions by viewing the problem as model fitting of a manifold in an abstract data space.We also derive a decision rule based on the geometric AIC.This rule can be used as a means of self-evaluation for testing if the robot motion is sufficient for structure-form-motion analysis without involving any empirically adjustable thresholds.To demonstrate this, we give examples using synthetic and real-image data.