Automatic calibration and 3D reconstruction for active robot vision

Visual sensing is important to robotics and automation. Efficient measurement and reconstruction of a 3D object requires a suitable vision system setting and view point planning strategy. In this talk, I will present our work on active vision and view planning using our camera-projector based vision system. With the recalibration method developed, the system’s settings can be on-line adjusted and its parameters can be recalibrated automatically whenever the configuration of the system is changed during a visual sensing task. Such recalibration requires only a single image of the scene, allowing the 3D measurement and reconstruction to be performed without interruption and making it adaptable to the environment. Using the active vision system, we investigated some issues in multi-view visual measurement, inspection and model acquisition tasks.