Within-class multimodal classification

Multimedia Tools and Applications - In many real-world classification problems there exist multiple subclasses (or clusters) within a class; in other words, the underlying data distribution is...     

Clustering semantically heterogeneous distributed aggregate databases

Databases developed independently in a common open distributed environment may be heterogeneous with respect to both data schema and the embedded semantics. Managing schema and semantic heterogeneities brings considerable challenges to learning from distributed data and to support applications involving cooperation between different organisations. In this paper, we are concerned mainly with heterogeneous databases that hold aggregates on a set of attributes, which are often the result of materialised views of native large-scale distributed databases. A model-based clustering algorithm is proposed to construct a mixture model where each component corresponds to a cluster which is used to capture the contextual heterogeneity among databases from different populations. Schema heterogeneity, which can be recast as incomplete information, is handled within the clustering process using Expectation-Maximisation estimation and integration is carried out within a clustering iteration. Our proposed algorithm resolves the schema heterogeneity as part of the clustering process, thus avoiding transformation of the data into a unified schema. Results of algorithm evaluation on classification, scalability and reliability, using both real and synthetic data, demonstrate that our algorithm can achieve good performance by incorporating all of the information from available heterogeneous data. Our clustering approach has great potential for scalable knowledge discovery from semantically heterogeneous databases and for applications in an open distributed environment, such as the Semantic Web.     

A scalable approach to integrating heterogeneous aggregate views of distributed databases

Aggregate views are commonly used for summarizing information held in very large databases such as those encountered in data warehousing, large scale transaction management, and statistical databases. Such applications often involve distributed databases that have developed independently and therefore may exhibit incompatibility, heterogeneity, and data inconsistency. We are here concerned with the integration of aggregates that have heterogeneous classification schemes where local ontologies, in the form of such classification schemes, may be mapped onto a common ontology. In previous work, we have developed a method for the integration of such aggregates; the method previously developed is efficient, but cannot handle innate data inconsistencies that are likely to arise when a large number of databases are being integrated. In this paper, we develop an approach that can handle data inconsistencies and is thus inherently much more scalable. In our new approach, we first construct a dynamic shared ontology by analyzing the correspondence graph that relates the heterogeneous classification schemes; the aggregates are then derived by minimization of the Kullback-Leibler information divergence using the EM (Expectation-Maximization) algorithm. Thus, we may assess whether global queries on such aggregates are answerable, partially answerable, or unanswerable in advance of computing the aggregates themselves.     

Improving angular error via systematically designed near-circular Gaussian-based feature extraction operators

In image filtering, the ‘circularity’ of an operator is an important factor affecting its accuracy. For example, circular differential edge operators are effective in minimising the angular error in the estimation of image gradient direction. We present a general approach to the computation of scalable circular low-level image processing operators that is based on the finite element method. We show that the use of Gaussian basis functions within the finite element method provides a framework for a systematic and efficient design procedure for operators that are scalable to near-circular neighbourhoods through the use of an explicit scale parameter. The general design technique may be applied to a range of operators. Here we evaluate the approach for the design of the image gradient operator. We illustrate that this design procedure significantly reduces angular error in comparison to other well-known gradient approximation methods.     

Adaptive vs. non‐adaptive strategies for the computation of optical flow

Confident adaptive algorithms are described, evaluated, and compared with other algorithms that implement the estimation of motion. A Galerkin finite element adaptive approach is described for computing optical flow, which uses an adaptive triangular mesh in which the resolution increases where motion is found to occur. The mesh facilitates a reduction in computational effort by enabling processing to focus on particular objects of interest in a scene. Compared with other state‐of‐the‐art methods in the literature our adaptive methods show only motion where main movement is known to occur, indicating a methodological improvement. The mesh refinement, based on detected motion, gives an alternative to methods reported in the literature, where the adaptation is usually based on a gradient intensity measure. A confidence is calculated for the detected motion and if this measure passes the threshold then the motion is used in the adaptive mesh refinement process. The idea of using the reliability hypothesis test is straightforward. The incorporation of the confidence serves the purpose of increasing the optical flow determination reliability. Generally, the confident flow seems most consistent, accurate and efficient, and focuses on the main moving objects within the image. © 2006 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 16, 35–50, 2006     

Fusion of Data and Knowledge for Safe UAV Landing

Autonomous Unmanned Aerial Vehicles (UAVs) have the potential to significantly improve current working practices for a variety of applications including aerial surveillance and search-and-rescue. However before UAVs can be widely integrated into civilian airspace there are a number of technical challenges which must be overcome including provision of an autonomous method of landing which would be executed in the event of an emergency. A fundamental component of autonomous landing is safe landing zone detection of which terrain classification is a major constituent. Presented in this paper is an extension of the Multi-Modal Expectation Maximization algorithm which combines data in the form of multiple images of the same scene, with knowledge in the form of historic training data and Ordnance Survey map information to compute updated class parameters. These updated parameters are subsequently used to classify the terrain of an area based on the pixel data contained within the images. An image''s contribution to the classification of an area is then apportioned according to its coverage of that area. Preliminary results are presented based on aerial imagery of the Antrim Plateau region in Northern Ireland which indicates potential in the approach used.     

Using evidence theory for the integration of distributed databases

Distributed databases allow us to integrate data from different sources which have not previously been combined. In this article, we are concerned with the situation where the data sources are held in a distributed database. Integration of the data is then accomplished using the Dempster–Shafer representation of evidence. The weighted sum operator is developed and this operator is shown to provide an appropriate mechanism for the integration of such data. This representation is particularly suited to statistical samples which may include missing values and be held at different levels of aggregation. Missing values are incorporated into the representation to provide lower and upper probabilities for propositions of interest. The weighted sum operator facilitates combination of samples with different classification schemes. Such a capability is particularly useful for knowledge discovery when we are searching for rules within the concept hierarchy, defined in terms of probabilities or associations. By integrating information from different sources, we may thus be able to induce new rules or strengthen rules which have already been obtained. We develop a framework for describing such rules and show how we may then integrate rules at a high level without having to resort to the raw data, a useful facility for knowledge discovery where efficiency is of the essence. © 1997 John Wiley & Sons, Inc.