Recognizing one-DOF industrial tools using invariant moments

In this paper, a simple but efficient approach is proposed to recognize one-DOF industrial tools. Since the shape is changed with the variation of the jaw angles and a feature vector obtained by conventional approach is not unique, we use the invariant moments and the ratio of area to perimeter squared of a boundary image to construct the required feature vector for object recognition. Two statistical classifiers based on the nearest-neighbor rule and the minimum-mean-distance rule are then utilized to pattern recognition. Experimental results show the good performance of this method in the noisy environment, as well as noise-free environment are also included.