Fusion of Multisensor Data

摘要

This paper treats two parts of the methodology of the fusion of multisensor data: conceptual and applied. In the concep tual part, we treat several basic questions including the fu sionability of various kinds of signals derived from the differ ent sets of raw data associated with separate sensor systems. We require that each derived signal has stand-alone signifi cance. We also derive some general results pertaining to the desirability of augmenting the set of measurements. In the applied part, the general methodology established above is employed in a problem in robotic acquisition involving the determination of the geometry (i.e., the elevation function) of an object resting upon a flat surface. In the model, two types of measurements are assumed: (1) a single acoustical pulse- echo scattering measurement in which the incident wave propagates downward, and (2) a set of optical measurements involving a downward-looking TV camera imaging the object under sequentially pulsed light sources. In the second type, the raw image data are assumed to be preprocessed according to the principles ofphotometric stereo. The two types of mea sured data were combined in accordance with the general concepts discussed in Part I of the paper. A computational example using synthetic data is presented and discussed.