Track-before-detect

Abstract: An on-board scanning or mosaic sensor is staring down from a satellite to a fixed ground point while producing a set of frames that contain target and background signals. Detecting dim moving targets should ideally be done by exhaustively searching over all in he maery(mached fltering), as ppoed possible trajectories in the imagery (matched filtering) as opposed to assembling trajectories from thresholded frames. The dynamic programming approach equivalently substitutes the above prohibitive exhaustive search by a feasible algorithm.

Abstract: The use of a non-co-operative television transmitter as the illuminator for a bistatic radar system is investigated. A signal processing scheme is developed that allows airborne targets to be detected and tracked using only the vision or sound carrier of the television broadcast. This scheme requires no synchronisation with the television transmitter, and uses the Doppler shift and bearing of target echoes to estimate the target's track. The signal processing scheme consists of a two-channel fast Fourier transform, to determine the Doppler and phase characteristics of the received signal, followed by time and frequency acting constant false alarm rate detection schemes to detect the target echoes. These echoes are corrected for phase errors arising from antenna element mutual coupling, and the corresponding bearings are calculated. A Kalman filter based tracking scheme is used to associate the individual Doppler and bearing returns belonging to different targets, resulting in Doppler and bearing profiles for each target. These are used to estimate the Cartesian co-ordinates and velocity of each target using an extended Kalman filter, initialised using a genetic algorithm and Levenberg-Marquardt optimiser. It is shown that targets can be detected and tracked over a large area, at ranges of up to 260 km.

Abstract: A typical sensor data processing sequence uses a detection algorithm prior to tracking to extract point-measurements from the observed sensor data. Track-before-detect (TkBD) is a paradigm which combines target detection and estimation by removing the detection algorithm and supplying the sensor data directly to the tracker. Various different approaches exist for tackling the TkBD problem. This paper compares the ability of several different approaches to detect low amplitude targets. The following algorithms are considered in this comparison: Bayesian estimation over a discrete grid, Dynamic Programming, Particle Filtering methods, and the Histogram Probabilistic Multi-Hypothesis Tracker. Algorithms are compared on the basis of detection performance and computation resource requirements.

Abstract: "Track-Before-Detect" (TBD) is a target tracking technique where no threshold is applied at each measurement frame. Instead, data are processed over a number of frames before decisions on target existence are made. The track is returned simultaneously with the detection. A simple algorithm is presented and demonstrated via simulations. A detailed analysis enables detection and tracking performance to be predicted for particular algorithm parameters. Good performance is observed at low signal-to-noise ratio (SNR), with rapid degradation as SNR is reduced further. For some cases the detection performance does not improve regardless of how many frames of data are processed. Tracking performance may also be poor even though detection performance is good.