Multitrack Fusion

Track fusion can be seen as part of sensor fusion, which is itself part of the even broader field of data fusion: "data fusion" combines digital measurements from sensors with symbolic information such as speakers' opinions, proposals, rules, texts in various languages and so on. Whatever the type of fusion, the objective remains the same: to make decision-making less risky, thanks to a function that is sufficiently automated to cope with the large, continuous flow of information from multiple sensors. If nothing were done, this deluge of data would overwhelm the operator (or even an entire team).

At first glance, it would seem logical to build this article by first describing the most general methods, i.e. those for information fusion, and then applying them to the specific case of lead processing. However, given the current state of knowledge in this dynamic and evolving field, such a top-down approach is not possible. Indeed, there is no such thing as a universal, unitary method of information fusion, and each field of application uses its own algorithms. This proliferation has several origins: the almost infinite number of possible combinations of sensors used for surveillance, the multiplicity of contexts of use linked to a wide variety of objectives, and lastly, the wide range of methods for taking account of uncertainty other than by means of classical probabilities. These "new" theories of uncertainty are mainly used in so-called "high-level" data fusion (artificial intelligence, expert systems, etc.).

It is for these reasons that this article presents the first level of fusion, that of tracks in surveillance systems, a problem whose solution requires few prerequisites: an engineer's current knowledge of detection and estimation theory is sufficient.

The first section presents the general organization of a surveillance system, in which the role of track fusion is, among other things, to provide, without delay, the best possible perception of the situation in the zone being monitored. The role of sensors (§ 2 ), tracking (§ 3