Tracking Process
Consolidation
CFAR
Association?
Prediction
Initiation
Update
Termination
Formation
(Track States)
Rate Monitoring
Control
Figure 1: Schematic workflow of automatic target tracking
Consolidation
CFAR
Association?
Prediction
Initiation
Update
Termination
Formation
(Track States)
Rate Monitoring
Control
Figure 1: Schematic workflow of automatic target tracking
Consolidation
Association?
Prediction
Formation
(Track States)
Rate Monitoring
Control
Figure 1: Schematic workflow of automatic target tracking
Tracking Process
The tracking process covers the processing of radar observations from the generation of individual target reports to the provision and management of the resulting tracks. New observations are associated with existing target tracks, their motion states are updated, and future target positions are predicted. Unassociated observations may be used to establish new tracks, while tracks without matching observations can initially be maintained and may eventually be terminated if suitable measurements remain unavailable. The block diagram in Figure 1 illustrates one possible schematic workflow.
Depending on the antenna rotation rate, pulse repetition frequency, and the dwell time of the radar beam on a target, several echo signals are usually received from the same target during one antenna scan (see dwell time and hits per scan in the chapter “Radar Basics”). The radar signal processor processes the received signals and uses suitable detection criteria to determine whether a possible target is present in a particular range cell. Several adjacent measurements or detections may belong to the same target. Before target tracking begins, this information must be combined into suitable observation reports.
Detection Consolidation
During detection consolidation, related detections, for example from adjacent range cells or successive transmitted pulses, are combined into a single observation report. This produces a representative target position from several individual measurements. Depending on the radar system, signal strength, echo extent, and additional characteristics may also be taken into account. Possible methods include:
- determining the target center by interpolation within a group of related detections;
- calculating a weighted average using signal strengths or other quality indicators;
- grouping detections according to additional properties, such as the Doppler spectrum.
The result is a plot, which serves as an observation report for the subsequent stages of automatic target tracking. Detection consolidation does not yet associate a plot with an existing track. It prepares the measurement data required for that operation.
Adaptive Detection Threshold (CFAR)
Radar echo detection is affected by noise, stationary clutter, precipitation, and other sources of interference. Under changing conditions, a fixed detection threshold may cause numerous false alarms or missed detections of weaker targets. Constant False Alarm Rate (CFAR) techniques therefore adapt the detection threshold to local interference conditions. Their objective is to maintain the specified false alarm probability as consistently as possible under the assumed interference statistics, rather than eliminating every false alarm.
CFAR processing belongs to radar detection and must be distinguished from the subsequent data association performed by the tracker. Whether particular measurements are consolidated before or after certain threshold decisions depends on the signal processing architecture. The block diagram therefore does not prescribe a mandatory processing sequence for every radar. Additional information about detection and false alarm probability is provided in the chapter “Radar Basics”.
Data Association (Track Association)
The plots supplied by radar signal processing must be associated with existing tracks. To accomplish this, the tracker uses previously estimated motion states to predict the expected target positions at the times of new measurements. Acceptance gates are established around these predicted positions. A plot within such a gate is considered a possible candidate for updating the corresponding track. If several plots fall within one gate, or a single plot falls within the gates of multiple tracks, additional criteria are required to determine the appropriate association.
Data association is therefore a separate decision-making operation: it determines which observation belongs to which target track. Subsequent state estimation, by contrast, determines how the new measurements are used to correct the previously estimated motion state. Difficult association situations may require methods that consider several possible assignments simultaneously, such as the Multiple Hypothesis Tracker.
Track Initiation
A single plot generally does not provide sufficient information to determine a target's motion reliably. Observation reports that cannot be associated with an existing track are therefore examined to determine whether they can establish a new target track. Several successive plots whose positions are consistent with plausible target motion may be used for this purpose. A newly initiated track may initially be regarded as tentative and is confirmed only after the specified criteria have been satisfied.
The number of observations required and the time interval over which they must occur depend on the radar system and its operational requirements. One commonly used criterion, for example, requires a specified minimum number of consistent observations within several successive antenna scans. This reduces the risk of creating false tracks from isolated false alarms. Further details are given under Track Initiation.
State Prediction and Track Updating (Track Prediction & Update)
For each existing track, the probable target position at the time of a new observation is predicted from the most recently estimated motion state. Position, velocity, and, where appropriate, acceleration may be combined into a state vector. The prediction also takes into account the selected motion model and the time elapsed since the last update.
Once a new plot has been associated with a track, its measured values are compared with the predicted state. A correction is calculated from the difference, resulting in an updated estimate of target position and motion parameters. For example, an α–β filter or a Kalman filter may be used for this purpose. These methods process measurements and predictions while accounting for their respective properties and uncertainties.
The calculation is repeated for each new suitable observation. The updated state estimate also provides the starting point for the next prediction. The mathematical principles are discussed in greater detail on the page about Track Smoothing and State Estimation.
Track Files and Track Updating
Information about tracked targets is stored in a track file or an equivalent data structure. For each active track, it contains the current estimated motion state and the management information required for further processing. Depending on the radar system, this information may include:
- track identifier, time stamp, and most recently associated measured position;
- estimated position, course, velocity, and, where applicable, acceleration;
- predicted target position at a later observation time;
- information about estimation accuracy and track quality or stability;
- additional measurement and target characteristics, such as signal strength or radial velocity;
- track confirmation status and information about previous updates.
When a new observation report is associated with an existing track, the corresponding state variables and management information are updated. The stored data may then be distributed to radar displays, air surveillance workstations, or other processing systems. In suitably equipped multifunction radars, they may also contribute to scheduling additional observations of selected targets.
Track Maintenance and Termination
If no suitable observation report is temporarily available for an existing track, this does not necessarily mean that the target has disappeared. The tracker may continue predicting the target's expected motion state and search for matching plots during subsequent observations. However, the uncertainty of the predicted position generally increases as more time passes since the last measurement. The acceptance gate may therefore need to be adjusted accordingly.
If suitable observations remain unavailable for an extended period, the reliability of the track decreases. Whether the track is maintained or terminated depends on specified criteria. These may include the number of missed updates, previous track quality, and the relevant operating conditions. The corresponding decisions are discussed in greater detail under Track Maintenance and Termination.
The tracking process thus combines radar detection and observation consolidation with data association, state estimation, prediction, and track management. Depending on the radar type, these functions may be organized differently or performed within a common processing unit. The essential objective is to transform individual observations, which may contain measurement errors, into target motion information that is as reliable and continuously updated as possible.