A Bore-sight Motion Detection Algorithm for Satellite Attitude Control
Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2007.
During an imaging pass of a remote sensing satellite, the satellite’s attitude has to be controlled so that the imager bore-sight sweeps out equal distances over time and so that images with a square aspect ratio are produced. The satellite attitude control system uses forward motion compensation (FMC) and time delay and integration (TDI) techniques to increase the quality of images. The motion of the scene relative to the satellite camera can be described by a two dimensional translation motion and a rotation about the camera bore-sight. This thesis describes an algorithm for measuring ground motion from viewfinder video data that can aid the satellite control system during imaging missions. The algorithm makes use of existing motion-from-video techniques – it operates in a hierarchical, feature-tracking framework. Features are identified on camera frames, and correspondences on consecutive frames are found by the Lucas and Kanade algorithm. A pyramidal image representation enables the estimation of large motions. The resultant sparse disparity map is used to estimate the three motion parameters, using a least squares fit to the projected motion equations. The algorithm was developed and implemented as part of the MSMI project. Results of tests carried out on simulated satellite viewfinder data (using the Matrix Sensor camera that was also developed for the MSMI project) confirms that the requirements are met.