Abstract
A binocular machine vision system having an autonomously moving active viewpoint is proposed. Using this active viewpoint, the system corresponds feature points between right and left images easily and calculates the spatial coordinates of features. When the system detects an interesting region in the camera's image data, the system moves the viewpoint to the region autonomously, and measures the position of feature points in the region. The interesting region for the system is defined to be a group of points which cannot be corresponded and/or exist periphery of the camera's view field. The image caught by a camera is affected by optical aberration of the camera. For this reason, measured coordinate of a feature point which exists far from the viewpoint cannot be reliable. Therefore the system incorporates a weighting processes to reduce the influence of unreliable measurement by varying the weight according to the distance from the viewpoint of the system. An experiment shows the effect of autonomous movement of viewpoint and weighting process on measuring the spatial coordinate of an object.
