Abstract
To realize autonomous robots for living support, object serving by robots using image processing is very useful. On this task, it must be needed to measure the exact three-dimensional position of the object in a real environment. However, in practice, object recognition by pattern matching is constrained by image changes because of lighting conditions, occlusions, object rotations and so on. In this research, we solve these problems by using SURF matching and decomposing homography, and then develop a whole system for object serving by an arm robot.