Showing papers by "Stephen J. Maybank published in 1997"

Vehicle tracking with applications to collision alert

Abstract: In this paper, the authors present an overview of a vision system that uses a camera mounted on a moving vehicle for locating, recognizing and tracking multiple vehicles. They first discuss how the background and local motions are determined. 3D model-based techniques are then used for obtaining information regarding temporal and spatial vehicle movements in order to detect potential hazards in time.

Path Prediction and Classification Based on Non-linear Filtering

Abstract: In path based filtering the usual state space is replaced by a finite dimensional approximation to the path space of the system. The information in the system model and the measurements is summarised by a probability density function on the approximating space. Path based filters are well suited to the inference of system behaviour over time. They have the advantage that the predicted or estimated paths are always physically plausible, in that they are realisations of the system model. The filter is applied to the model based tracking of cars. Measurements of the position and orientation of a moving car are obtained by fitting a wire frame model to a sequence of video images. The filter estimates the velocity, acceleration and steering angle of the car. Experiments show that the steering angle can be estimated after tracking for 1 s and the acceleration after 2 s.