IST Lunch Bunch

The Co-registration of Optically Sensed Images and Correlation (COSI-Corr) is a software package developed at the California Institute of Technology for accurate geometrical processing of optical satellite and aerial imagery. First released to the academic community in 2007, COSI-Corr is used for a wide range of applications in Earth Sciences, which take advantage of the software capability to register, with very high accuracy, images taken from different sensors and acquired at different times.
In this lecture, we will review the traditional COSI-Corr processing chain to quantify sub-pixel displacement between time-series of images, and we present the latest modules which aim at analyzing time-series of stereoscopic imagery to document 3D variations of the ground surface. In particular, these latest improvements include an N-image bundle block adjustment to jointly optimize the viewing geometry of multiple acquisitions, an improved multi-scale image matching based on Semi-Global Matching (SGM) to extract high resolution topography, and the triangulation of multi-temporal disparity maps to derive 3D ground motion.
Numerous applications drawn from recent studies will be presented, such as the study of ground motion induced by recent and past earthquakes, the measurement of 3D ice flow, the tracking sand dunes on Mars, or the production of high resolution digital terrain models to quantify building damages after large disasters.
Finally, we will highlight the potential for applying these techniques on a large scale to provide global monitoring of our environment.

Joint work with François Ayoub, Jiao Lin, Bruno Conejo, and Jean-Philippe Avouac.