Semi-automatic extraction and mapping of dyke swarms based on multi-resolution remote sensing images: Applied to the dykes in the Kuluketage region in the northeastern Tarim Block

Abstract

The geometry and spatial distribution of mafic dyke swarms provide information about geodynamics. However, mapping large-scale dyke swarms manually is a time-consuming and expensive process. Here we present an efficient method to map dyke swarms semi-automatically by using multi-resolution remote sensing images. This method is applied in the Kuluketage region in the northeastern Tarim Block, northwest China where mafic dykes of multiple trends are widely exposed and with the dominant NW trend being of probable Neoproterozoic age. The tracing of dykes, including their orientation, length, thickness, density, and distribution is done by integrating 30-m multispectral Landsat 8 and 2-m panchromatic CORONA KH-4B imagery. The Gramm-Schmidt Spectral Sharpening approach is used to generate color composite and fusion images with 2-m resolution. The Canny edge detector and Hough transform are used to detect dyke edges. Combining the detection of dyke edges with ArcGIS™, the dyke orientation, length, density, thickness, and exposed area are measured. The accuracy of the semi-automatic method exceeds 75% in comparison with manual mapping of the dykes in the Kuluketage region.