A Fuzzy Shapes Characterization for Robotics

Abstract

We proposed [9] an original membership function building for fuzzy pattern recognition. This method uses only one shape feature — the compactness — leading to a simple and fast determination of membership function. We improved this method adding the possibility of control for the slopes of functions and unifying their representation [10,11]. We implemented this method in machine vision area, where for the robot’s eye it must process a continuos image data series in real time. In the reason to reduce the need of preliminary human inspection, we proposed to add a new feature — the elongatedness — to be used in conjunction with the membership function built from compactness [12]. In this paper we propose a unified membership function, built by using the compactness and the elongatedness features of shapes together.