Three-dimensional surface reconstruction based on a self-organizing feature map

Abstract

We present a new surface-reconstruction method based on Kohonen's algorithm for a self-organizing feature map. The primary goal of our investigation is to build a general and robust system, which can cope with multiple sensor information. We try to address viewpoint independence, fusion of data from various sources and noiserobustness. The key idea is to let the self-organizing feature map define a surface parameter grid which is most appropriate. With the help of results, which were achieved with the German space-tele- robotic experiment ROTEX as background scenario, the performance of the described method can be assessed.