Merging data

Data fusion is a recent technique for processing multi-source information. It makes it possible to merge multiple data concerning a phenomenon, in order to obtain a more well-founded judgement than that obtained from all the sources considered in isolation.

The very term data fusion was introduced in the USA in the 70s, and data fusion techniques underwent extensive development in the 80s, which is still ongoing. This meteoric rise is due to a combination of two phenomena: advances in computer technology and urgent military requirements. It should be noted that, today, applications have spread beyond this military concept, as fusion is proving necessary in applications such as robotics, transport and image processing.

Authors usually distinguish between sensor fusion and data fusion.

The first involves combining information from different sensors, such as radar, infrared or visible cameras, laser rangefinders, laser imagers, ultraviolet sensors, ultrasonic sensors, seismic sensors, etc. Data fusion has a wider spectrum of uses, since it is concerned with combining all sources of knowledge, whether from sensors, navigation systems, various databases (cartographic data, documentation, digital terrain models, know-how rules) or even previous analyses or fusions.

This article presents the problems posed, in both contexts, and an overview of the methods that can be used to solve them.