Advanced visual surveillance systems not only need to track moving objects but also interpret their patterns of behaviour. This means that solving the information integration problem becomes very important. We use conceptual knowledge of both the scene and the visual task to provide constraints. We also control the system using dynamic attention and selective processing. Bayesian belief network (BBN) techniques support this as well as allowing us to model dynamic dependencies between parameters involved in visual interpretation. We illustrate these arguments using experimental results from a traffic surveillance application. In particular, we show that using expectations of object trajectory, size and speed for the particular scene can improve robustness and sensitivity in dynamic tracking and segmentation. We also show that behavioural evaluation under attentional control can be achieved using a combination of a static BBN tasknet and dynamic network (DBN). The causal structure of the...

Robust tracking and segmentation of faces is a prerequisite for face analysis and recognition. In this paper, we describe an approach to this problem which is well suited to surveillance applications with poorly constrained viewing conditions. It integrates motion-based tracking with modelbased face detection to produce segmented face sequences from complex scenes containing several people. The motion of moving image contours was estimated using temporal convolution and a temporally consistent list of moving objects was maintained. Objects were tracked using Kalman filters. Faces were detected using a neural network. The essence of the system is that the motion tracker is able to focus attention for a face detection network whilst the latter is used to aid the tracking process. 1 Introduction In order to analyse and recognise peoples' faces in realistically unconstrained environments, robust tracking and segmentation is a prerequisite. This provides a sequence of face images normalise...

We are entering an era of more intelligent cognitive vision systems. Such systems can analyse activity in dynamic scenes to compute conceptual descriptions from motion trajectories of moving people and the objects they interact with. Here we review progress in the development of flexible, generative models that can explain visual input as a combination of hidden variables and can adapt to new types of input. Such models are particularly appropriate for the tasks posed by cognitive vision as they incorporate learning as well as having sufficient structure to represent a general class of problems. In addition, generative models explain all aspects of the input rather than attempting to ignore irrelevant sources of variation as in exemplar-based learning. Applications of these models in visual interaction for education, smart rooms and cars, as well as surveillance systems is also briefly reviewed.

In this paper we show how a Pearl Bayes network of inference can be used with a GIS in order to combine information from different sources of data for the purpose of classification. Data may include satellite images, topographic maps, geological maps etc, each one with its own resolution and accuracy. We show how this uncertainty in the input data is incorporated in the network and develop also a method to construct the conditional probability matrices used by the network. We demonstrate our approach within the framework of the problem of assessing the risk of desertification of some burned forests in the Mediterranean region. 1

This paper reviews research that addresses the problems of extracting descriptions of object behaviour from image sequences. Vision systems are now capable of delivering trajectory-based descriptions of moving objects in a scene but little work has been done on the higher-level spatio-temporal reasoning needed for the computation of behavioural descriptions. This level of understanding, which allows meaningful descriptions of what is happening in a scene, seems to be a prerequisite for communication between users and machine based vision systems. The approaches discussed here can be separated into three main classes: those that treat the problem as an off-line query-based process, those that attempt an on-line model-based interpretation, and those that adopt a more active vision stategy. Some evidence from the psycholinguistic literature, event perception, and recent developments in reactive planning are brought together to support the proposal that active, purposive frameworks are req...

In this dissertation we present an architecture to incorporate collateral information in the high level computer vision task of object location. The declarative specification of the scene hypothesis and the domain independent control algorithm that exploits spatial information to drive the vision process, are two major contributions of this dissertation. This work has been theoretically grounded in constraint satisfaction algorithms. Apart from the application of standard CSP algorithms, several new techniques in constraint satisfaction have been developed. We provide an algorithm that performs consistency filtering for certain types of non-binary constraints. We also presente an innovative application of interval arithmetic in constraint satisfaction. Though interval constraint satisfaction is an extensive research area, we believe this to be the first attempt at using interval arithmetic to reduce the domain generation cost of finite domain CSPs. A new algorithm has been proposed her...