Proceedings of the Fifth International Conference on Information Fusion. FUSION 2002. (IEEE Cat.No.02EX5997)

In support of the AFOSR program in Information Fusion, the CNS Technology Laboratory at Boston University is developing and applying neural models of image and signal processing, pattern learning and recognition, associative learning dynamics, and 3D visualization, to the domain of Information Fusion for Image Analysis in a geospatial context. Our research is focused by a challenge problem involving the emergence of a crisis in an urban environment, brought on by a terrorist attack or other man-made or natural disaster. We aim to develop methods aiding preparation and monitoring of the battlespace, deriving context from multiple sources of imagery (high-resolution visible and low-resolution hyperspectral) and signals (GMTI from moving vehicles, and ELINT from emitters). This context will serve as a foundation, in conjunction with existing knowledge nets, for exploring neural methods in higher level information fusion supporting situation assessment and creation of a common operating picture (COP).

There is an obvious need to be able to integrate both linguistic-based and stochastic-based input information in data fusion. In particular, this need is critical in addressing problems of track association, including cyber-state intrusions. This paper treats this issue through a new insight into how three apparently distinct mathematical tools can be combined: "boolean relational event algebra" (BREA), "one point random set coverage representations of fuzzy sets" (OPRSC), and "complexity-reducing algorithm for near optimal fusion" (CRANOF).

Bayesian networks are an attractive modeling tool for human sensing, as they combine an intuitive graphical representation with efficient algorithms for inference and learning. Temporal fusion of multiple sensors can be efficiently formulated using dynamic Bayesian networks (DBNs) which allow the power of statistical inference and learning to be combined with contextual knowledge of the problem. Unfortunately, simple learning methods can cause such appealing models to fail when the data exhibits complex behavior We first demonstrate how boosted parameter learning could be used to improve the performance of Bayesian network classifiers for complex multimodal inference problems. As an example we apply the framework to the problem of audiovisual speaker detection in an interactive environment using "off-the-shelf" visual and audio sensors (face, skin, texture, mouth motion, and silence detectors). We then introduce a boosted structure learning algorithm. Given labeled data, our algorithm modifies both the network structure and parameters so as to improve classification accuracy. We compare its performance to both standard structure learning and boosted parameter learning. We present results for speaker detection and for datasets from the UCI repository.

We propose a new fusion algorithm based on the Dempster-Shafer theory of evidence. The main interest of this method is a new distribution of mass functions. Generally the methods used are the mass consonant distribution and the partially mass consonant distribution. The originality of this work is to define uncertain and inaccurate data by using a fuzzy statistical classification algorithm such as FSEM (Fuzzy Stochastic Estimation Maximization). Application to multiband image fusion produces interesting results for classification.

This paper presents a new and simple approach to the problem of multiple sensor data fusion. We introduce an efficient algorithm that can fuse multiple sensor measurements to track an arbitrary number of objects in a cluttered environment. The algorithm combines conventional Kalman filtering techniques with probabilistic data association methods. A Gauss Markov process model is assumed to handle sensor outputs at various sampling frequencies and random nonequidistant time intervals. We applied the algorithm to post-process the digital range returns of radar sensors to improve their quality. Since the static noise returns have near-zero velocity, the algorithm associates a certain track with each digital return, and estimates the track velocity, thereby allowing for removal of false returns originating from static pattern noise.

This paper explores the subject of Cramer-Rao lower bounds for unbiased sensor registration algorithms. Two Cartesian coordinate systems are considered: a two-dimensional regional plane for tactical surveillance and the Earth-centred-Earth-fixed (ECEF) coordinates for wide area surveillance. The theoretical performance bound is of fundamental importance both for algorithm performance assessment and for the prediction of best achievable performance given target and sensor location, and number and accuracy of measurements. The performance of the iterative least-squares registration algorithm, obtained by Monte Carlo simulations, is compared to the theoretical bounds.

Traditionally, radar or sensor performance is reviewed in terms of detection ranges and measurement accuracy. Today, when the radar often is a part of a locally integrated sensor system, which shall cooperate with other sensor systems, other performance parameters are required for evaluation and comparison of various solutions. The present paper gives an analysis method for descriptions of the tracking performance over the entire surveillance volume. It is also shown how the tracking performance can be visualized and interpreted. Furthermore, given that the user has chosen his preferred performance metrics, the paper shows how the analysis can be used as a decision aid, by comparing the existing system performance with the performance of a planned system configuration. Finally, the advantages of sensor data fusion are often asked for to be quantified. With the approach taken this can be done. The paper shows the tracking performance for a system of sensors, at each point in the volume, either as picking the best set of data from the individual sensors, or as the optimally fused data.

The paper combines IMM and JPDA for tracking of multiple possibly maneuvering targets in case of clutter and possibly missed measurements while avoiding sensitivity to track coalescence. The effectiveness of the filter is illustrated through Monte Carlo simulations.

Decision making is always a difficult assignment for everyone in every context and more information eases the process and makes it more accurate. More data does not always mean more information. Rough sets alone can be adopted to handle hierarchical structure in the family of criteria using decision rules based on preference model. The rules are induced from the preferential information given by the Decision Maker (DM) in the form of examples of decisions. The existing rough set techniques are applicable in information systems where condition and decision attributes are distinguishable. But, in general, the information are available in the form of a data table, called an information system or knowledge representation system, where rows are labeled by alternatives and columns by attributes. In such cases, the condition and decision attributes are not distinguishable. The present paper offers a solution for making decisions in an efficient way when enough data are available, relevant or irrelevant. In a Multiple Criteria Decision Analysis (MCDA) problem, determination of weights is an important aspect of the Analytic Hierarchy Process (AHP). Based on rough set concepts, a model, viz., the interval AHP model with interval data, for determination of the weights has been considered here. According to the necessary and possibility concepts of rough set, we obtain the lower and upper evaluation models, respectively. After obtaining the local interval weights, a method is proposed for calculating the global weights, and also for ranking the alternatives. The proposed method will be applicable for an information system as well as for a decision table.

This paper proposes a new scheme of fusing cortex transform and brightness based features obtained by local windowing operation. Energy features are obtained by applying popular cortex transform technique within a sliding window rather than the conventional way, while we define three features namely directional surface density (DSD), normalised sharpness index (NSI), and normalized frequency index (NFI) as measures for pixel brightness variation. Fusion by simply vector tagging as well as by correlation is performed in the feature space and then classification is done using minimum distance classifier on the fused vectors. It is interesting that the brightness features, though inferior on some natural images, often produces smoother texture boundary in mosaic images, whereas energy features show the opposite behavior. This symmetrically inverse property is combined through vector fusion for robust classification of multi-texture images obtained from Brodatz album and VisTex database. Classification outcome with confusion matrix analysis shows the robustness of the scheme.