|M.Sc Student||Zhao Jingjing|
|Subject||Incorporation of Thermal Infrared Imaging into Spatial|
3-D Scene Modeling
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Sagi Filin|
|Full Thesis text|
Three-dimensional thermography offers a powerful technology to capture detailed information of surface temperature of objects and environments. It benefits a wide variety of applications including agricultural, industrial, medical, search and rescue, and building energy leakage detection, to name only a few examples. Nonetheless, current approaches seeking to establish extraction of geometric content, e.g., 3-D thermography, suffer from pre-processing related drawbacks. These include, complicated procedures for radiometric and metric calibration, low resolution images and low image quality and texture. Aiming for a viable framework for thermal 3-D processing, this thesis proposes a pipeline involving the geometric image calibration and subsequent enhancement and matching procedures aiming towards 3-D reconstruction of thermographic data. Focusing on applicable solution, the research introduces a novel approach for geometric calibration of such image data. Whereas existing approaches require expensive calibration targets and elaborate setups, we show that a low-cost solution can be obtained to calibrate the camera intrinsic parameters in a computational efficient manner and without compromising accuracy. For the purpose of enhancing image contrast, our image enhancement algorithm in thermal modality proves to achieve better than the conventional approach. Finally, focusing on feature extraction and correspondences, a comprehensive evaluation of the performance of popular feature detection and descriptor generating algorithms leads to deeper understanding on image matching and reconstructing leading to 3-D thermography modeling.