Localisation, Back-Projection and Fusion of LWIR Hyperspectral Data from FTIR Imaging Sensors


Ressource 1Download: [PUBLIC]Thesis_report_Ayoub-Fatihi.pdf (3723.60 [Ko])
State: Public
Version: After imprimatur
License: All rights reserved
Serval ID
A Master's thesis.
Localisation, Back-Projection and Fusion of LWIR Hyperspectral Data from FTIR Imaging Sensors
Fatihi Ayoub
Yaagoubi Reda
Sebari Imane, Lorenz Sandra, Thiele Sam
Institution details
Institut Agronomique et Vétérinaire Hassan II
Publication state
Issued date
The use of high-resolution images and data acquired from terrestrial or aerial platforms enables detailed examination of features on the earth’s surface, such as mineral deposits, geomorphological formations, and geological structures. This information is used to map and study the subsurface, understand geodynamic processes, and assess environmental impact. Hyperspectral remote sensing uses the electromagnetic spectrum reflected or emitted by objects on the Earth’s surface for various applications, including mineral identification and analysis of the distribution and properties of materials. However achieving an accurate representation of the observed target requires the fusion and co-registration of data obtained from various sensors due to the different acquisition methods of each portion of the electromagnetic spectrum.
This study presents a semi-automatic co-registration workflow for the integration of geological close-range remote sensing data. The method is designed to overcome the challenges of hypercloud formation, where multiple sources of hyperspectral data are collected and integrated in 3D space. The methodology involves a camera calibration process to eliminate initial distortions, followed by stitching of small Long-Wave Infrared hypercubes obtained from a Hyper-cam device to form a hyper-mosaic. The next step involves matching and transformation to establish an affine transformation between the hyper-mosaic and an RGB image, its quality was evaluated by comparing the differences between the coordinates of selected points and their corresponding transforms. Next a Structure-from-Motion workflow is carried out to generate a point cloud and the determination of the RGB frame’s camera position and orientation. This then allows back-projection can be performed onto the point cloud to generate a hypercloud. The back-projection process was evaluated through the calculation of the three-dimensional distance between selected points and their respective projection. An average discrepancy of 3.56 cm was achieved. The hypercloud can be filled with data from other regions of the electromagnetic spectrum and/or used to produce mineral maps and other geologically relevant products.
co-registration, LWIR, hypercloud, Telops hyper-cam, SfM, homography, affine transformation
Create date
01/03/2024 15:40
Last modification date
01/03/2024 15:46
Usage data