Information Fusion in Signal and Image Processing: Major Probabilistic and Non-Probabilistic Numerical Approaches

Chapter 11

Fusion of Non-Simultaneous Elements of Information: Temporal Fusion

11.1. Time variable observations

Temporal fusion of data allows us to handle instantaneous information, i.e. information whose value is only valid at a given time. This fusion is used when the parameters obtained from observation vary with time.

The problem therefore is to track the evolution of these observations, as opposed to an overall characterization of the evolution. The following example illustrates this instantaneous nature: when observing a beating heart, the temporal fusion of data makes it possible to estimate the instantaneous blood flow. With a non-temporal fusion, it would be possible to estimate the mean blood flow over a long period of time. On the other hand, temporal fusion would be needed to observe the evolution of the mean blood flow depending on, for example, the evolution of the patient's level of stress. This shows that we always have to define concepts related to time such as “moment”, “evolution”, “period of time” according to the application because a length of time can be “long” in certain cases, or “short” or even “negligible” in others.

Thus, temporal fusion involves the evolution or the modification of data:

– it may involve the modification of the object being observed. For example, when analyzing the heart's movement, focus will be directed to its morphological modifications over time;

– it may involve the relative evolution of two objects. In robotic manipulation, for ...