Active 3D recognition system based on fourier descriptors

This paper presents a new 3D object recognition/pose strategy based on reduced sets of Fourier descriptors on silhouettes. The method consists of two parts. First, an off-line process calculates and stores a clustered Fourier descriptors database corresponding to the silhouettes of the synthetic model of the object viewed from multiple viewpoints. Next, an on-line process solves the recognition/pose problem for an object that is sensed by a real camera placed at the end of a robotic arm. The method avoids ambiguity problems (object symmetries or similar projections belonging to different objects) and erroneous results by taking additional views which are selected through an original next best view (NBV) algorithm. The method provides, in very reduced computation time, the object identification and pose of the object. A validation test of this method has been carried out in our lab yielding excellent results.