| 摘要 |
A method for target-tracking of objects via observations from a sensor, wherein a geographical region within which at least one object is present is divided into a predetermined number of states, and the geographical region is divided into subregions so that adjacently disposed subregions all contain a share of states in common. Target-tracking is initiated in a first subregion, whereupon a first observation is associated with the object, and the probable state of the object is estimated. For each new time interval, a new observation is chosen for which a new state is estimated, whereupon the estimated sequence of states in the subregion is updated. When target-tracking is begun in a new subregion, the state history from the preceding subregions is transferred to the new adjacent subregion so that the state estimate in the new subregion is based on the state history in the preceding adjacent subregions. A transition is initiated when hte probability that the object has moved to a new subregion exceeds a certain predetermined value.
|
