Image Segmentation for Hydrothermal Alteration Mapping Using PCA and Concentration–Area Fractal Model

Abstract

Information extraction from processed remotely sensed images, in the case of missing initial spectra of pixels, can be a challenge for the users. In such situations, application of conventional methods based on spectral properties of pixels is impractical. We took advantage of the fractal theory for image segmentation of a principal component (PC) image for hydrothermal alteration mapping. The selected input images included short wave infrared bands of ASTER imagery covering the Darrehzar porphyry copper mine and surrounding areas with well-known hydrothermal alteration zones. Hydrothermal alteration like other geological processes can show spatial distribution with fractal properties. Principal component analysis was used to enhance hydrothermal alteration associated with the Darrehzar porphyry copper deposit. None of the resulting PCs were appropriate to portray clearly important alteration types in the study area. The PC1 image, which contains more than 98% of variance of the input bands, was selected for image segmentation using a digital number–area technique based on the established concentration–area fractal model. This technique was proposed based on frequency distributions and spatial correlation and variability of pixel values. The resulting hydrothermal alteration map indicates intense phyllic, weak phyllic, and propylitic as the main alteration types exposed at the surface of the Darrehzar area. In addition, the proposed technique delineated the phyllic zone in the exposed mine pit and a transition zone between inner phyllic and surrounding propylitic alteration zones. Field investigation and sampling in 23 locations including spectral measurements, XRD and thin section studies, confirmed the accuracy of the classified image by the proposed technique.