Robust Range Data Segmentation using Geometric Primitives for Robotic Applications

G. Taylor and L. Kleeman (Australia)


range images, segmentation, geometric primitives


This paper presents a data-driven approach for segmenting range data to enable a humanoid robot to perform inter active domestic tasks. Range data is segmented into ge ometric primitives (planes, spheres, cylinders and cones), allowing the robot to recognize many simple objects. The algorithm calculates initial segments using depth disconti nuities, creases and changes in surface type, with subse quent merging steps to correct for over-segmentation. We develop a novel surface type classification method using analysis of the Gaussian image and convexity of surface patches. Our method eliminates the need for choosing an arbitrary approximating function as required by other sur face type classifiers. We also present novel techniques to estimate initial parameters for fitting geometric primi tives to segmented regions. Experimental results using real range data demonstrate the effectiveness of our techniques.

Important Links:

Go Back