CGI Unveils New Stochastic Imaging Method to Improve Subsurface Accuracy

CGI is proud to highlight a new contribution in geophysical imaging published in the journal Inverse Problems. Co-authored by Patrick Belliveau and Eldad Haber, the paper, titled "Parametric level-set inverse problems with stochastic background estimation," addresses a common industry challenge: reconstructing underground structures when the surrounding background medium is heterogeneous or poorly understood.

The team developed a sophisticated algorithm using sample average approximation and accelerated stochastic gradient descent to minimize the "noise" caused by uncertain background data. This allows for far more precise shape recovery of mineral deposits or reservoirs. By treating the background as a Gaussian random field, the method effectively filters out environmental uncertainty that often leads to imaging errors in traditional models.

This breakthrough was successfully tested on a demanding three-dimensional inverse conductivity problem, proving its scalability for large-scale geophysical projects. For our clients in mining, oil and gas, and geothermal exploration, this research translates into more reliable subsurface maps and reduced drilling risk in complex terrains. At CGI, we continue to integrate these cutting-edge mathematical frameworks into our consultancy services to deliver superior imaging results.