Automatic extraction of lineament information from satellite images using digital elevation data

Using conventional visual interpretation in lineament analysis presents two main problems. The first is subjectivity, introduced because of the bias of various interpreters. The second problem is that lineaments detected from satellite images are constrained by the direction of the illumination source. Since lineament identification mainly involves recognition of diagnostic morphological features, the use of digital elevation can contribute significant information about these features. Further, in generating images using digital elevation data, the direction of illumination can easily be controlled. Thus, the use of digital elevation data offers the possibility of revealing features not apparent in regular satellite images.We discuss a sequential line detection method for extraction of linear features from digital elevation data. In this method, raw elevation data is used for generating shaded relief images using the Lambertian reflection model, wherein the illumination direction is controlled by the user. The Directional Segment Detection Algorithm (DSDA) is used for detecting linear topographic features in user-defined trends. Locational information about these linear features is stored in the computer as coordinate pairs amenable to editing and subsequent analysis. Finally, three-dimensional terrain models are generated by combining the digital elevation data and satellite images. The experiments were carried out using digital elevation data of southwest Japan and Landsat MSS data.