Imaging Subsurface Cavities From Microgravity Data Using Hopfield Neural Network
Ahmed S.K. Salem, Eslam A. Elawadi and Keisuke Ushijima
Kyushu University, Japan.
Contacts: salem@mine.kyushu-u.ac.jp (Ahmed S.K. Salem)
Abstract
The presence of subsurface cavities may cause ground subsidence problems that affect environmental, military, and cultural structures. Imaging of these cavities from microgravity survey data become a common application in the recent years. However, there is an increasing demand for automatic interpretation techniques that could locate cavities from the microgravity data in the field. In this paper, we have developed a new method for automatic detection of subsurface cavities from microgravity data using a Hopfield neural network. We assumed that the observed gravity anomaly is produced by an equivalent source (cylinder or sphere), which has an amplitude factor related to the radius, density contrast and depth. The Hopfield network is used to optimize the amplitude factor of the equivalent source at a set of regular locations. For each location, the Hopfield network reaches its stable energy state. The location of the cavity corresponds to the location yielding the minimum Hopfield energy. The method was tested using synthetic and field data of subsurface cavities buried at different depths. In all cases, the method can estimate the cavities locations with a good success.
raeg2003@tansa.kumst.kyoto-u.ac.jp
Last modified: Thu Oct 03 18:52:45 2002