Sequential track extraction

Iterative tracking algorithms must be initiated appropriately. Under simple conditions, this is not a difficult task, as has been shown above (Eq. 3.8). For low observable objects, i.e. objects embedded in a high clutter background [1-5] or in case of incomplete measurements [6, 7], more than a single set of observations at particular data collection times are usually necessary for detecting all objects of interest moving in the sensors' fields of view. Only then, the probability density iteration can be initiated based on 'extracted' object tracks, i.e. by tentative tracks, whose existence is 'detected' by a detection process working on a higher level of abstraction.