Novel Signal Processing Techniques For The Exploitation Of Thermal Hyperspectral Data

This doctoral thesis attemps to propose a novel signal processing chain, aimed to exploit data acquired by long wave infrared (LWIR) hyperspectral sensors. In the LWIR, infrared radiation from an object is directly related to its temperature, i.e. hotter the surface, higher the emitted thermal energy. Hyperspectral sensors capture the radiated energy from the objects (target) in a large number of consecutive spectral bands within the LWIR, e.g. with the aid of a prism, in order to estimate the spectrum (spectral emissivity) and the temperature of the surface material. In this framework, two main challenging tasks affect the development and the deployment of thermal hyperspectral sensors: - atmospheric correction: the process of estimate and compensate the thermal radiation produced by the atmosphere, that affects the thermal radiation procuded by the target. This process is made more complicated by the complex combination of atmospheric parameters, such as temperature, pressure, water vapor concentration, aerosol concentration etc., and by the complex profile that such a parameters follow from ground towards the space, passing through all the atmospheric layers. - temperature emissivity separation: often abbreviated as TES problem; once the atmospheric correction of the signal has been applied, it represents the task of jointly estimate the surface spectral emissivity and temperature of the observed target. The estimate represents an ill-posed problem: given the thermal radiation of the target, measured at N different channels of the deployed hyperspectral sensor, the unknows of the problem are N + 1, i.e. the Nc spectral sample of its emissivity, and the temperature. In this framework, the thesis aim to provide a processing chain for the exploitation of the thermal radiation, captured by hyperspectral sensors, focused on the task of target detection, i.e. recognize a given material within a complex remote sensing scenario. The applications of such a signal processing chain may be the more disparate, from environmental remote sensing (climate change, volcanos monitoring, gas and waste pollution etc.) to defence and surveillance purposes.