FAST OBJECT LOCALIZATION WITH REAL TIME 3D LASER RANGE SENSOR SIMULATION

Abstract

Abstract: 3D laser range sensors provide range data of objects in a scene. Most of the known approaches for object recognition and localization in range data try to extract features and determine the position of the object from this information. In opposite we create a virtual scene, make a virtual range image with a virtual 3D laser range sensor and compare the result with range data of the real 3D laser range sensor. This idea is integrated in a system for automated robotic bin picking. Our simulation process is combined with a hierarchical registration process using the iterative closest point algorithm and Progressive meshes to refine our simulation results. 1.