LINEAMENT ANALYSIS OF SATELLITE IMAGES USING A SEGMENT TRACING ALGORITHM (STA)

Regional fault patterns contribute important information for prelimina ry exploration for mineralized areas and characterization of geologic structure. Faults may correspond with lineaments appearing on remotely sensed images. In this paper, a new lineament identification method, namely the Segment Tracing Algorithm (STA), is proposed, and its appli cation in mountainous areas of southwestern Japan is discussed. The pr inciple of the STA is to detect a line of pixels as a vector element b y examining local variance of the gray level in the digital image, and to connect retained line elements along their expected directions. Th e threshold values for the extraction and the linkage of line elements are direction dependent. The advantages of the proposed method over u sual filtering methods are its capability to (1) trace only continuous valleys and (2) extract more lineaments that parallel the sun's azimu th and those located in shadow areas. A LANDSAT Thematic Mapper (TM) s ubscene was used for the analysis. The band 4 subscene of the LANDSAT TM image was determined to be appropriate for the lineament analysis b y comparing several images with a shaded-relief image derived from a d igital elevation model (DEM).