Detection and localization of aerosol releases from sparse sensor measurements

Abstract

In this thesis we focus on addressing two aspects pertinent to biological release detection. The first is that of detecting and localizing an aerosolized particle release using a sparse array of sensors. The problem is challenging for several reasons. It is often the case that sensors are costly and consequently only a sparse deployment is possible. Additionally, while dynamic models can be formulated in many environmental conditions, the underlying model parameters may not be precisely known. The combination of these two issues impacts the effectiveness of inference approaches. We restrict ourselves to propagation models consisting of diffusion plus transport according to a Gaussian puff model. We derive optimal inference algorithms utilizing sparse sensor measurements, provided the model parameterization is known precisely. The primary assumptions are that the mean wind field is deterministically known and that the Gaussian puff model is valid. Under these assumptions, we characterize the change in performance of detection, time-to-detection and localization as a function of the number of sensors. We then examine some performance impacts when the underlying dynamical model deviates from the assumed model. In addition to detecting an abrupt change in particles in an environment, it is also important to be able to classify the releases as not all contaminants are of interest. For this reason, the second aspect of addressed is feature extraction, a stage where sensor measurements are reduced to a set of pertinent features that can be used as an input to the classifier.

(cont.) Shift invariance of the feature set is critical and thus the Dual Tree Complex Wavelet Transform (DT CWT) is proposed as the wavelet feature domain.