Guang Jin

Date of Award


Level of Access

Open-Access Dissertation

Degree Name

Doctor of Philosophy (PhD)


Spatial Information Science and Engineering


Silvia E. Nittel

Second Committee Member

Michael F. Worboys

Third Committee Member

M. Kate Beard-Tisdale


Today, technology developments enable inexpensive production and deployment of tiny sensing and computing nodes. Networked through wireless radio, such senor nodes form a new platform, wireless sensor networks, which provide novel ability to monitor spatiotemporally continuous phenomena. By treating a wireless sensor network as a database system, users can pose SQL-based queries over phenomena without needing to program detailed sensor node operations. DBMS-internally, intelligent and energyefficient data collection and processing algorithms have to be implemented to support spatial query processing over sensor networks. This dissertation proposes spatial query support for two views of continuous phenomena: field-based and object-based. A field-based view of continuous phenomena depicts them as a value distribution over a geographical area. However, due to the discrete and comparatively sparse distribution of sensor nodes, estimation methods are necessary to generate a field-based query result, and it has to be computed collaboratively ‘in-the-network’ due to energy constraints. This dissertation proposes SWOP, an in-network algorithm using Gaussian Kernel estimation. The key contribution is the use of a small number of Hermite coefficients to approximate the Gaussian Kernel function for sub-clustered sensor nodes, and processes the estimation result efficiently. An object-based view of continuous phenomena is interested in aspects such as the boundary of an ‘interesting region’ (e.g. toxic plume). This dissertation presents NED, which provides object boundary detection in sensor networks. NED encodes partial event estimation results based on confidence levels into optimized, variable length messages exchanged locally among neighboring sensor nodes to save communication cost. Therefore, sensor nodes detect objects and boundaries based on moving averages to eliminate noise effects and enhance detection quality. Furthermore, the dissertation proposes the SNAKE-based approach, which uses deformable curves to track the spatiotemporal changes of such objects incrementally in sensor networks. In the proposed algorithm, only neighboring nodes exchange messages to maintain the curve structures. Based on in-network tracking of deformable curves, other types of spatial and spatiotemporal properties of objects, such as area, can be provided by the sensor network. The experimental results proved that our approaches are resource friendly within the constrained sensor networks, while providing high quality query results.