The Kalman filter and all its simplified or sophisticated versions are usually the basic estimation technique used to develop different association mechanisms. Based on the predicted target state it becomes possible to estimate the expected observation position and to correlate the arriving target returns and false measurements with tracks. Five basic association techniques are considered in this chapter: the Nearest and Strongest Neighbor approaches (effective just for sparse environments), the Probabilistic and Joint Probabilistic Data Association Filters, and finally, the Multiple Hypothesis Tracking algorithm. The PDA/ JPDA algorithms are considered in [1] as a special case of MHT. Their calculations are similar to those, required for MHT. According to it, the advantage of the JPDA is that it is a relatively simple recursive method, which does not require the storage of past observation data nor multiple hypotheses. An apparent disadvantage associated with the PDA/JPDA is the lack of an explicit mechanism for track initiation. As discussed in [1], however, it is suitable to employ other batch- type algorithms for track initiation and then to use the PDA/JPDA for track maintenance. It is also noted that the most important factor in the choice of MHT versus JPDA methods is probably the FA density. For high FA densities, such as in sonar or radar air-to-ground tracking applications, MHT is considered not feasible, and the JPDA is favored. The MHT is considered feasible for the lower false target densities associated with the radar air-to-air tracking problem. In contrast to this opinion, it is shown in [4], that introducing batch-processing track initiation procedure based on application of Hough Transform it becomes possible to decrease significantly the required computational load, generally improving the MHT algorithm performance.
展开▼
