Research paper A new artefacts resistant method for automatic lineament extraction using Multi-Hillshade Hierarchic Clustering (MHHC) Jakub Šilhavý a,n , Jozef Minár b , Pavel Mentlík c , Ján Sládek d a Section of Geomatics, Department of Mathematics, Faculty of Applied Sciences, University of West Bohemia in Pilsen, Univerzitní 22, Pilsen 306 14, Czech Republic b Faculty of Natural Sciences, Comenius University, Bratislava, Slovak Republic c Centre of Biology, Geoscience and Environmental Education, Faculty of Education, University of West Bohemia, Pilsen, Czech Republic d Department of Physical Geography, Geomorphology and Natural Hazards, Institute of Geography, Slovak Academy of Sciences, Štefánikova 49, Bratislava 814 73, Slovak Republic article info Article history: Received 26 August 2015 Received in revised form 25 March 2016 Accepted 28 March 2016 Available online 31 March 2016 Keywords: Lineaments GIS Automatic extraction Bohemian forest Central Western Carpathians abstract This paper presents a new method of automatic lineament extraction which includes the removal of the ‘artefacts effect’ which is associated with the process of raster based analysis. The core of the proposed Multi-Hillshade Hierarchic Clustering (MHHC) method incorporates a set of variously illuminated and rotated hillshades in combination with hierarchic clustering of derived ‘protolineaments’. The algorithm also includes classification into positive and negative lineaments. MHHC was tested in two different territories in Bohemian Forest and Central Western Carpathians. The original vector-based algorithm was developed for comparison of the individual lineaments proximity. Its use confirms the compatibility of manual and automatic extraction and their similar relationships to structural data in the study areas. & 2016 Elsevier Ltd. All rights reserved. 1. Introduction Lineaments are generally considered to be linear features manifesting in the land surface and land cover reflecting dis- continuities of geological structures (mainly faults). Although various phenomena can form lineaments (rock boundaries, sedi- mentary layers, wetness and vegetation changes – see e.g. Gupta, 2003), distinct linear landforms are most frequently used to ex- tract geological structures (Smith and Clark, 2005; Smith and Wise, 2007; Evans, 2012). If lineaments detection is based ex- clusively on the morphometric properties of the land surface, then the lineaments can be termed ‘morpholineaments’ (Minár and Sládek, 2009). Although morpholineaments are automatically ex- tracted either directly from a Digital Elevation Model (DEM) (Vaz, 2011; Mallast et al., 2011) or from different derived surfaces, e.g. second derivatives of DEM (Wladis, 1999), shaded relief (hillshade) is the most frequently used derived surface (Abdullah et al., 2010; Masoud and Koike, 2011a; Jordan and Schott, 2005). Image pre-processing (edge enhancement, noise removal using thresholding) followed by edge linking methods (Hough Trans- form, Canny edge detector) are mostly used for automated lines extraction (Table 1). In some cases, the pre-processing is part of the extraction (closed-source software modules). Morpholineaments can be considered not only as a surface ex- pression of particular lithospheric faults, joints and lithological boundaries (e.g. Solomon and Ghebreab, 2006; Štěpančíková et al., 2008; Batayneh et al., 2012), but also as an expression of a mor- photectonic field – a manifestation of lithospheric stress fields in the landforms (Urbánek, 2005; Minár and Sládek, 2009; Sládek, 2010). When producing a morphotectonic field model, even a small artificial misrepresentation of the morpholineaments direction (artefacts) can lead to problematic interpretations of results. Artefacts formation during a raster based analysis is pointed out and solved in this paper. The main objective of this paper is to present a new Multi- Hillshade Hierarchic Clustering (MHHC) artefacts resistant method for automated lineaments extraction. The second goal is to eval- uate the correlation between automatically and manually deli- neated lineaments, test the algorithm's ability to detect linear geological features (such as faults and linear parts of rock boundaries) and extract the main tectonically significant direc- tions for their following evaluation in morphotectonic analysis. Although subjective visual assessment is the most common approach for validation of extracted lineaments (Kageyama and Nishida, 2004; Jordan and Schott, 2005), more objective ap- proaches have been published. For example, Abdullah et al. (2010) computed simple statistics of count and length of lineaments to Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/cageo Computers & Geosciences http://dx.doi.org/10.1016/j.cageo.2016.03.015 0098-3004/& 2016 Elsevier Ltd. All rights reserved. n Corresponding author. E-mail address: jakub.silhavy@gmail.com (J. Šilhavý). Computers & Geosciences 92 (2016) 9–20