Experimental investigation of line-of-sight motion reconstruction from satellite images

The objective of this thesis project is to design and size a test bench for evaluating the performances of an optical image stabilization system intended for Earth observation payloads and to evaluate the performances of motion estimation algorithms to measure the motion of the target being captured by a camera. Indeed, the stabilization system is designed to improve image quality by compensating in real time and coherently with respect to the target motion. The first phase of the project consisted of an extensive research about the importance of image stabilization for Earth observation payloads and its implementation in past missions. A methodology was found to quantitatively and qualitatively assess the utility of the proposed method, followed by a trade-off analysis to select optimal components for simulating the relative motion between the sensor and the target, supported by worst-case scenario calculations. In the second part of the thesis, different algorithms for the estimation of the spacecraft line-of-sight motion from images of the Earth have been implemented and tested. Four algorithms have been selected for their accuracy, computational efficiency, and robustness under various noise scenarios, namely Lucas-Kanade, Horn-Schunck, 1D Projection and Phase correlation. The algorithms have been tested on an experimental set-up, adapted from the one previously designed to comply with the hardware components available in the laboratory.