Over the last decades strong efforts have been made to apply new spaceborn technologies to the study and possible forecast of strong earthquakes. In this study we use ASTER/TERRA multispectral satellite images for detection and analysis of changes in the system of lineaments previous to a strong earthquake. A lineament is a straight or a somewhat curved feature in an image, which it is possible to detect by a special processing of images based on directional filtering and or Hough transform. "The Lineament Extraction and Stripes Statistic Analysis" (LESSA) software package, developed by Zlatopolsy (1992, 1997). We assume that the lineaments allow to detect, at least partially, the presence ruptures in the Earths crust, and therefore enable one to follow the changes in the system of faults and fractures associated with strong earthquakes. We analysed 6 earthquakes occurred in the Pacific coast of the South America and XXX with the Richter scale magnitude >4.5. They were located in the regions with small seasonal variations and limited vegetation to facilitate the tracking of features associated with the seismic activity only. It was found that the number and orientation of lineaments changed significantly about one month before an earthquake approximately, and a few months later the system returns to its initial state. This effect increases with the earthquake magnitude, and it is much more easily detectable in case of convergent plate boundaries (for example, Nasca and South American plates). The results obtained open a possibility to develop a methodology able to evaluate the seismic risk in the regions with similar geological conditions.