This article presents a pedestrian tracking methodology using an infrared sensor for surveillance applications. A distinctive feature of this study compared to the existing pedestrian tracking approaches is that the road network information is utilized for performance enhancement. A multiple model particle filter, which uses two different motion models, is designed for enabling the tracking of both road-constrained (on-road) and unconstrained (off-road) targets. The lateral position of the pedestrians on the walkways are taken into account by a specific on-road target model. The overall framework seamlessly integrates the negative information of occlusion events into the algorithm for which the required modifications are discussed. The resulting algorithm is illustrated on real data from a field trial for different scenarios.

Using prior information about the road network will improve the estimation performance for a road constrained target significantly. Several estimation methods have been proposed to handle the multi-modality that arise in a road target tracking application. One popular filter suitable for this kind of non-linear problems is the Particle Filter, but a major drawback is that the Particle filter requires a large amount of particles as the state dimension increases to maintain a good approximation of the filtering distribution. In this paper a Rao-Blackwellized Particle Filter based approach is proposed to reduce the dimension of the state space in road target tracking applications. Furthermore, it is also shown how prior information about the probability of detection can be used to improve the estimation performance further.

Airborne surveillance systems equipped with a vision/infrared camera require good knowledge about the position and orientation of the camera for successful tracking of ground targets. In particular, this is essential when incorporating prior information, like road maps, that is expressed relative a global reference system. Usually, it is possible to obtain good positioning with inertial/satellite navigation systems, but estimating the orientation is generally more difficult. It might be possible to use SLAM (Simultaneous Localization and Mapping) or image registration approaches to support the navigation system, but not always since such approaches require stable features in the images. In this paper the problem of simultaneous orientation error estimation and road target tracking is considered by assuming that the target is constrained to a known road network. A particle filter approach is proposed and it is shown that the result of this filter is close to the performance of the ideal case where the orientation error is perfectly known. However, the performance depends on how informative the road path is and in rare cases the orientation error is unobservable.

This work considers a sensor management problem where a number of road bounded vehicles are monitored by a UAV with a gimballed vision sensor. The problem is to keep track of all discovered targets and simultaneously search for new targets by controlling the pointing direction of the vision sensor and the motion of the UAV. A planner based on a state-machine is proposed with three different modes; target tracking, known target search, and new target search. A high-level decision maker chooses among these sub-tasks to obtain an overall situational awareness. A utility measure for evaluating the combined search and target tracking performance is also proposed. By using this measure it is possible to evaluate and compare the rewards of updating known targets versus searching for new targets in the same framework. The targets are assumed to be road bounded and the road network information is used both to improve the tracking and sensor management performance. The tracking and search are based on flexible target density representations provided by particle mixtures and deterministic grids.

In this work we consider a target tracking scenario where a moving observer with a bearings-only sensor is tracking a target. The tracking performance is highly dependent on the trajectory of the sensor platform, and the problem is to determine how it should maneuver for optimal tracking performance.The problem is considered as a stochastic optimal control problem and two sub-optimal control strategies are presented based on the Information filter and the determinant of the information matrix as the optimization objective. Using the determinant of the information matrix as an objective function in the planning problem is equivalent to using differential entropy of the posterior target density when it is Gaussian. For the non-Gaussian case, an approximation of the differential entropy of a density represented by a particle mixture is proposed. Furthermore, a gradient approximation of the differential entropy is derived and used in a stochastic gradient search algorithm applied to the planning problem.

Exploration is in this work defined as the task of efficient information gathering of areas, building, roads, etc., by controlling a pan/tilt camera on a sensor platform. Good exploration is characterized by several images from different directions of the areas of interests and that the images can be used to create maps, video mosaics and multi-view imagery for anomaly and change detection. In this paper an aerial exploration framework based on the information filter is presented. The work is inspired by research on optimal trajectory for bearings-only tracking. A number of static grid points represent the area to be explored and the problem is to plan the trajectory of the sensor platform and the pointing direction of the camera to maximize the exploration performance of the grid points.