Search

In this paper we address the issue of mobility in geosensor networks, inspired by the computational challenges imposed by modern surveillance applications. More specifically we consider networks of optical sensors (video and still cameras), and present a spatiotemporal framework for the management of information captured in them. In this context, mobility is addressed at two levels, considering mobile objects in the area monitored by a network, and mobile sensors observing such objects. Our interest lies on the data acquisition and storage problems that arise in this setting. We identify certain key issues behind the development of a general framework for knowledge acquisition and data storage in geosensor networks, namely: spatiotemporal object modeling; similarity metrics to compare spatiotemporal objects; storing and indexing spatiotemporal objects in a geosensor network; and network management using spatiotemporal techniques. We present some emerging approaches that address these key issues and thus outline a general framework for information and sensor management in mobile sensor networks.

Granularity in time and space has a fundamental role in our perception and understanding of various phenomena. Currently applied analysis methods are based on a single level of granularity that is user-driven, leaving the user with the difficult task of determining the level of spatiotemporal abstraction at which processing will take place. Without a priori knowledge about the nature of the phenomenon at hand this is often a difficult task that may have a substantial impact on the processing results. In light of this, this paper introduces a spatiotemporal data analysis and knowledge discovery framework, which is based on two primary components: the spatiotemporal helix and scale-space analysis. While the spatiotemporal helix offers the ability to model and summarize spatiotemporal data, the scale space analysis offers the ability to simultaneously process the data at multiple scales, thus allowing processing without a priori knowledge. In particular, this paper discusses how scale space representation and the derived deep structure can be used for the detection of events (and processes) in spatiotemporal data, and demonstrates the robustness of our framework in the presence of noise.

A common approach in database queries involves the multi-dimensional representation of objects by a set of features. These features are compared to the query representation and then combined together to produce a total similarity metric. In this paper we introduce a novel technique for similarity learning within features (attributes) by manipulating fuzzy membership functions (FMFs) of different complexity. Our approach is based on a gradual complexity increase adaptable to problem requirements. The underlying idea is that less adaptable functions will act as approximations for more complex ones. We begin by interpolating a set of planes in the training dataset and due to linearity we get a fast first impression of the underlying complexity. We proceed to interpolate two asymmetrical sigmoidal functions whose initial approximations are calculated from the plane properties. If satisfactory accuracy is not achieved we provide advanced modeling capabilities by investigating FMFs parameters and convolving their output with additional functions.

In this paper, we present a differential change-oriented model for the storage and communication of spatio-temporal information. The focus is on the development of model primitives and operators to support the aggregation of change over time and the propagation of change across resolution. Our investigation is motivated by recent advances in image-based geospatial databases, with constantly increasing update frequencies, and diverse user communities performing queries of various levels of detail. The primitives and operators presented here extend existing qualitative operators to support the management of quantitative and geometric information within a change-oriented spatio-temporal environment. We also show how the design of our model results from ‘change semantics’ at different granularities. By doing so, advanced communication operations can be addressed within our model in an efficient way.

An approach to semi-automated linear feature extraction from aerial imagery is introduced in which Kohonen's self-organizing map (SOM) algorithm is integrated with existing GIS data. The SOM belongs to a distinct class of neural networks which is characterized by competitive and unsupervised learning. Using radiometrically classified image pixels as input, appropriate SOM network topologies are modeled to extract underlying spatial structures contained in the input patterns. Coarse-resolution GIS vector data is used for network weight and topology initialization when extracting specific feature components. The Kohonen learning rule updates the synaptic weight vectors of winning neural units that represent 2-D vector shape vertices. Experiments with high-resolution hyperspectral imagery demonstrate a robust ability to extract centerline information when presented with coarse input.

This paper addresses the topic of semi-automatic object extraction for GIS data capture. It includes novel algorithms and conceptual strategies, together with performed experiments, encountered problems and adopted solutions. In particular, we present an algorithm and obtained results for semi-automatic extraction of road networks from SPOT imagery using wavelet-transformed images and cost functions expressing local gray value variations and global continuity constraints. For larger scale images and for various object types, we present our technique of least squares template matching for edge extraction, which uses local gray value variations to precisely identify edge locations. Finally, we propose a global approach for semi-automatic object outline detection, whereby least squares matching provides the mathematical foundation, while global continuity is enforced through the introduction of object-type-dependent shape constraints.