Sensor registration is an important problem that must be considered when attempting to perform any kind of data fusion in multimodal, multisensor target tracking. In this multiple target tracking (MTT) application, any inaccuracies in the registration can lead to false tracks being created, and tracks of true targets being stopped prematurely. This article introduces a method for simultaneously tracking multiple targets in a surveillance region and estimating appropriate sensor registration parameters so that sensor fusion can be performed accurately. The proposed method is based around particle belief propagation (BP), a recent but highly efficient framework for tracking multiple targets. The proposed method also uses a hierarchical model which allows for multiple processes to be linked and interact with one another. We present a comprehensive set of simulations and results using differing, asynchronous sensor setups, and compare with a random finite set (RFS) approach, namely the sequential Monte Carlo (SMC)-probability hypothesis density (PHD) filter. The results show the proposed method is 17% more accurate than the RFS approach on average.

The number of nodes in sensor networks is continually increasing, and maintaining accurate track estimates inside their common surveillance region is a critical necessity. Modern sensor platforms are likely to carry a range of different sensor modalities, all providing data at differing rates, and with varying degrees of uncertainty. These factors complicate the fusion problem as multiple observation models are required, along with a dynamic prediction model. However, the problem is exacerbated when sensors are not registered correctly with respect to each other, i.e., if they are subject to a static or dynamic bias. In this case, measurements from different sensors may correspond to the same target, but do not correlate with each other when in the same Frame of Reference (FoR), which decreases track accuracy. This paper presents a method to jointly estimate the state of multiple targets in a surveillance region, and to correctly register a radar and an Infrared Search and Track (IRST) system onto the same FoR to perform sensor fusion. Previous work using this type of parent-offspring process has been successful when calibrating a pair of cameras, but has never been attempted on a heterogeneous sensor network, or in a maritime environment. This article presents results on both simulated scenarios and a segment of real data that show a significant increase in track quality in comparison to using incorrectly calibrated sensors or single-radar only.

Sensor registration is fundamental in sensor fusion. Inaccuracies in sensor location and rotation can manifest themselves into the measurements used in Multiple Target Tracking (MTT), and dramatically degrade its performance. These registration parameters are often estimated separately to any multitarget estimation, which could lead to increased computational expense, and also to systematic errors. Recent works have shown that MTT algorithms derived from Belief Propagation (BP) are computationally efficient and highly scalable for large tracking scenarios. This work presents a hierarchical Bayesian model inspired by single-cluster methods from the Random Finite Set (RFS) literature, that allow for the registration parameters to be estimated jointly with the multiple target tracking. Simulations are carried out on a multistatic radar network containing two radars with a relative range and azimuth bias between them. Results are presented for a particle-BP MTT algorithm, and it's performance is compared to that of a Sequential Monte Carlo (SMC)-Probability Hypothesis Density (PHD) filter. The results show that the BP algorithm outperforms the PHD implementation in terms of accuracy by around 10%.

Modern sensor platforms carry an increasingly diverse range of sensors onboard, in order to estimate target positions inside their common surveillance region. Off-the-shelf sensors often provide measurements at different rates, and with different and potentially varying levels of uncertainty. Heterogeneous and asynchronous sensor networks make the sensor fusion problem more challenging as multiple measurement models and a dynamic prediction model are required. Moreover, a key challenge to address in fusion systems is that of sensor bias. Any relative bias between sensors could result in measurements not correlating with one another in a common frame of reference, and therefore vastly reducing track accuracy. This work presents novel results on a joint method that estimates both external angular bias between a radar and a camera, and the states of multiple targets. The proposed technique uses a particle-based implementation of Belief Propagation (BP), and compares with Random Finite Set (RFS)-based approaches. Initial results show that the BP approach outperforms the RFS approaches in terms of accuracy by around 50% when using the Optimal Sub-Pattern Assignment (OSPA) metric.

Target tracking in radar uses mature methods that have been around for many years. The Extended Kalman Filter has long been the typical choice of tracking filter in radar systems due to its low computational effort and its simple implementation. With many recent advances in hardware and processing capability on airborne platforms, it is now becoming possible to integrate higher-complexity algorithms into real-time systems. This work presents an alternative method of tracking a single target using a hybrid filtering approach with an alternative state space parametrization. This approach will be compared to an EKF, UKF and a particle filter using simulated data and root mean square error.

The necessity for maintaining surveillance in airborne environments is ever growing. Criminals and terrorists are finding new and elaborate means of attack, and small UAVs such as quadcopters and hexacopters could be a possible threat. Their small size and agile movement will make them difficult to detect. This work aims to determine whether or not these small UAVs can be detected at short range using radar, and if so, track them over time using a suitable filter such as a Bernoulli filter.