As mentioned in the previous section, to put a cubic object passing through a square hole, the robot first needs to estimate the orientation of the object and then turns it to the same orientation as the hole. Provided Assumption (A1), a well-known algorithm relying only upon visual information to achieve this purpose is based on the image moments [22]. Fig. 5(a) illustrates the orientation of a cubic object defined in this study (the angle θc in the figure). According to [22], the image-moments-based approach estimates the angle θc by the orientation of an equivalent ellipse with the same twoorder centered moments as the object. This approach works well for an arbitrary object; however, it may fail to estimate the orientation of a regular polygon object (i.e., the cubic object considered here) since the corresponding equivalent ellipse is close to a circle. This problem highlights the importance of orientation estimation problem for the regular polygon objects. This section hence presents a novel image-based object orientation estimation algorithm that efficiently and accurately estimates the orientation of a geometric polygon object from a segmented image.