Tracking involves both data association and the process of filtering, smoothing, or predicting. Data association involves determining the origin of the measurements (i.e., determine whether a return is a false alarm, clutter, or a valid target and assess which returns go with which tracks or is this the first return from a given target).

Given that the return is properly associated, an algorithm is needed to include this latest measurement in a manner that will improve the estimate of the next expected position of the target. Early trackers, such as the alpha-beta-gamma filter, used precomputed fixed gains that were sometimes changed based on maneuver detection.

They were simple to code and required small amounts of memory and throughput. As tracking advanced, radars began to use the EKF.Many filter states were used in ballistic missile defense (BMD) and early warning trackers.

More modern tracking approaches use non uniformly scheduled pulses, Kalman filtering of multiple sensors, nonlinear filters, interacting multiple model (IMM), joint probabilistic data association (JPDA), and multiple hypothesis tracking (MHT).

Decision-directed techniques, such as MHT, can result in a growing memory that must be pruned as possibilities are deemed unlikely. As target density or clutter increases, many false tracks can initiate but over time it becomes obvious which are actual and which are bogus.

The Hough transform can be used to track/detect straight line trajectories or those generalized for curvature. Phased arrays provide flexibility for minimizing the energy to track targets with a given accuracy or impact point prediction (IPP).

Options include revisit interval, dwell time, and beam width spoiling. A tracking radar can use a high prf that avoids range blindness on the tracked target while providing ample Doppler space free of clutter.

If a search radar is ambiguous in range, a different prf must be used on each dwell to resolve range ambiguities. If a target is within the unambiguous range interval, the range cell where the detection occurs does not change. If the target is beyond the range ambiguity distance, the range cell number changes due to the range fold over.

The Chinese remainder theorem can be used to unravel the true range based on several ambiguous range measurements. The range estimate, however, can be grossly in error if an unambiguous range cell number is off by a single range cell due to measurement noise.

Other approaches of resolving range ambiguities avoid this problem. For example, the entire instrumented range can be laid out for each dwell with a return placed at all corresponding ambiguous range cells.

By summing the dwells in each range cell, the one with the highest count will be the true range since they all occur in this cell. To prevent errors due to a slight range error, one can sum both the range cell and its adjacent neighboring cells over all dwells.

This will ensure that slight misses will be correctly counted in the summation. Methods that resolve range ambiguities for a point target are not very effective for a weather radar where the target is distributed. Multiple targets can produce ghosts when unraveling unambiguous ranges.

No comments:

Post a Comment

Related Posts Plugin for WordPress, Blogger...