Mattias Larsson and Max Modig present their Master's Thesis on Tuesday 4/6 at 10.00 in MH:333

Abstract:

With a growing global demand for improved surveillance and security, the thermal camera has seen its use cases and market share expand. Unlike the visual camera, the thermal camera does not require external illumination. Instead, it relies on the fact that all matter with a temperature above absolute zero emits thermal radiation. The thermal camera measures intensity, but another characteristic of electromagnetic radiation is its polarization state. The polarization state can provide additional information about the scene, such as the orientation and shape of objects and their material composition. By utilizing the features of the polarized domain, polarimetry has been used to solve a wide range of problems in a number of different fields. This project has revealed another application of polarimetric imaging in the area of license plate recognition. A full system, including a polarized thermal camera and required software solutions, has been constructed for this purpose. The limitations and restrictions of polarimetric imaging and how it specifically relates to the problem of license plate recognition have also been investigated. The observed polarization effect in the LWIR domain is mainly influenced by the temperature difference between the object and its optical background, the incidence angle of the radiation, and the refraction index of the object. These three factors are, in turn, influenced by other, more or less controllable aspects. Through visual evaluation, it is clear that the polarimetric images provide additional and relevant information that is not present in a pure thermal image. Further evaluation of the implemented system, given that proper conditions are met, confirms an empirical advantage. The implemented system serves as a proof of concept and requires more development before any real-world implementations can be considered.