3P1-I03 Deformable Object Classification with Decision Tree Learning using Local Normal Feature and Application to the Washing Transportation Task(Life management and robotics)

Abstract

In daily life, we have a large number of deformable objects. Robot need to handle those things in ohter way unlike undeformable objects, because those will change their shapes easily and it isn't simple to recognize their shapes or structure. Therefore, Robot should recognize whether the target object is deformable or not. In this paper, in order to classify object as deformable or not, we focus on deformable objects' 3d point cloud wrinkles feature. We test some features which are calculated from object surface normal and we found the average of FPFH is the best feature to classify softness of object.