Shape Recognition by Cable Model for Cables of Multiple Types and its Application to Optimal Holding Position

Abstract

A novel method of shape representation through the cable model is proposed for cables of various types used in many industrial production lines. Principal component analysis is properly utilized in the method for detecting length orientations of objects. Segmentation is a fundamental process which enables to handle complex shaped cables as a train of linearly approximated simple elements. The cable model for shape recognition is represented by calculated segment center vectors and outer edge vectors. Through utilization of the cable model, position of optimal holding for robot arm is estimated. Then, experiments of holding and shape recognition show the availability and effectiveness of the proposed method.