Extended Object Tracking Using Monte Carlo Methods

Abstract

This correspondence addresses the problem of tracking extended objects, such as ships or a convoy of vehicles moving in urban environment. Two Monte Carlo techniques for extended object tracking are proposed: an interacting multiple model data augmentation (IMM-DA) algorithm and a modified version of the mixture Kalman filter (MKF) of Chen and Liu [1], called the mixture Kalman filter modified (MKFm). The data augmentation (DA) technique with finite mixtures estimates the object extent parameters, whereas an interacting multiple model (IMM) filter estimates the kinematic states (position and speed) of the manoeuvring object. Next, the system model is formulated in a partially conditional dynamic linear (PCDL) form. This affords us to propose two latent indicator variables characterizing, respectively, the motion mode and object size. Then, an MKFm is developed with the PCDL model. The IMM-DA and the MKFm performance is compared with a combined IMM-particle filter (IMM-PF) algorithm with respect to accuracy and computational complexity. The most accurate parameter estimates are obtained by the DA algorithm, followed by the MKFm and PF.

Contextual Mixture Tracking

“…To deal with multiple objects within a tractable configuration space, we adopt the mixture model proposed in [21] to factorize the configuration space according to the number of targets: (1) where is the total number of objects, and the mixture weights satisfy . In Sequential Monte Carlo framework (also known as Particle Filter), the filtering distribution are approximated by a set of weighted particles :…”

“…We use the position information of each proposed particle to calculate the distance from the groundtruth position to the predicted position and sum over the distances of all particles to represent the prediction precision of the proposal distribution. We compare CPD with commonly used Gaussian prior [19], [1] on the sequence. The result is shown in Fig.…”

“…Most of them can be categorized into two classes. The first category is termed as joint trackers [1], [4], [18], which straightforwardly extends the motion state space by concatenating the states of all targets (so all targets' motions are dependent of each other). This state space will expand exponentially with the number of targets.…”

Track-before-detect procedures for detection of extended object

“…For example, a high-resolution radar provides a useful measure of down-range extent given a reasonable signal-to-noise ratio (SNR). The possibility to additionally make use of the high-resolution measurements is referred as extended object tracking [1]. Estimation of the object shape parameters is especially important for track maintenance [2] and for the object type classification.…”

“…More recent approaches to tracking extended targets have been investigated by assuming that the measurements of target extent are available [1][2][3][4][5]. However, the measurements of extended targets provided by the high resolution sensor are inaccurate in a low SNR environment since those are obtained by threshold-based decisions made on the raw measurement at each scan.…”