Automatic lineament extraction from potential-field images using the Radon transform and gradient calculation

Abstract

Linear anomalies are critical in the interpretation of gravity and magnetic data. Visual identification of lineaments is usually done by experienced interpreters, and identification results then have to undergo a digitization or import procedure. The traditional identification method has unavoidable subjectivity and inefficiency. To overcome these limitations, we fuse the Radon transform (RT) with gradient calculation to process gravity or magnetic data and to realize automatic detection and extraction of lineaments. As part of the detection procedure, we define the RT-based mean gradient (MG), effective mean gradient (EMG), and residual mean gradient (RMG) in order to highlight long linear segments or to enhance short linear ones in the transform domain. The gradient forms are applied self-adaptively and self-selectively to gravity or magnetic images according to specific conditions. Gradient directions are also taken into account in the transformation procedure to emphasize the characteristics of linear anomalies. To extract the position and length of the detected lineaments from the transform domain, a constraint inverse searching method (CISM) is given and used to locate the starting and end points of the lineaments. The method can deal with the condition that there is at least one linear section in a specific direction or that separate linear sections may belong to one lineament. Through tests with synthetic images and with real data from the Haijiao upheaval area in the East China Sea Basin, the detection and extraction methods are shown to be more effective and robust than the conventional RT applications. The results from the real data roughly coincide with major geologic faults that are visually identified. These results show that the methods constitute a useful tool to aid fault interpretation.