Skip to Main Content
We present a method to aggregate and classify the 2D outline of arbitrary shaped objects from 3D-range data. As a mobile robot only captures a limited part of an object in a single measurement, the raw data of several measurements are combined to build a data set as comprehensive as possible. Doing so the classification uses all previous measurements. As the classification uses local curvature based features it can classify partial and complete shapes. The approach combines a spline approximation method with a scalable method for shape registration, combination and classification. Experimental results illustrate the ability of combining and classifying real world objects on a mobile robot.