Range image registration using plane extraction by the CAN2

This paper describes range image registration to fuse three-dimensional surfaces of range images taken from around an object. By means of using the competitive associative net called CAN2 for plane extraction, we constructed two methods: one is for the case where the planes of the floor and the wall are available, and the other is for the case where the available planes are on the floor and the object. With experimental results using the real images obtained by the laser range finder (LRF), we examine the performance of the methods, and present several problems to be solved in future research studies.