In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
OBJECT-ORIENTED METHODS FOR LANDSLIDES DETECTION USING HIGH 
RESOLUTION IMAGERY, MORPHOMETRIC PROPERTIES AND 
METEOROLOGICAL DATA 
I. Sandric a,b ’ *, B. Mihai 3 , Z. Chitu 3 , A. Gutu b , I. Savulescu 3 
a University of Bucharest, Faculty of Geography, Bdul. Nicolae Balcescu, No.l, Sect. 1., 010041, Bucharest, Romania - 
sandricionut@yahoo.com 
b ESRI Romania SRL, Roma Street, No 8, Sect. 1, 011774, Bucharest, Romania 
Commission VI, WG VI/4 
KEY WORDS: Landslides, Identification, Modelling, Recognition, Object, Pattem, Segmentation 
ABSTRACT: 
Mapping landslides and building landslides inventory have received a special attention from a wide range of specialist. In building a 
landslide inventory an important step is the spatial delineation of the landslides body, followed by the landslides classification 
according with an international used classification system and the identification of other landslides characteristics. The main methods 
for landslides mapping are based on field observation, image interpretation and stereo-restitution. Our paper discusses a semi- 
automated process based on objected-oriented analysis for landslides bodies’ delineation. Several recent papers Moine et al. 2009, 
Tapas et al. 2010 had similar approaches for landslides bodies’ delineation and classification using objected-oriented analysis 
combined with spectral and morphometric properties of the landslides. Our approach is similar with Tapas et al. 2010, but we take 
into account, besides the morphometric properties, the meteorological data for the periods when the landslides have occurred. The 
algorithm is using high resolution aerial images with a spatial resolution 0.5 meters, a DEM with a spatial resolution of 2.5 meters 
and daily meteorological data for the year 2005. The meteorological data was spatial interpolated and the images were used in the 
objected oriented analysis and this has led to a significant increase in the number of corrected indentified landslides. The algorithm 
was tested in the administrative area of Breaza Town from Romanian Curvature Sub- Carpathians, for which a detailed landslides 
inventory was available 
1. INTRODUCTION 
1.1 Aim of the paper 
Landslides are frequent phenomenon for the Subcarpathians 
regions of Romania. Lithological features based mainly on clay 
and marl combined with permeable rocks, the torrential rainfall 
regime as well as a strong human impact has a key role for the 
landslide related hazards (Balteanu, 1986, 1997; Ielenicz et al., 
1999,2002). 
The landslide-related hazards are a real environmental problem 
within the local and regional development of Subcarpathian 
areas, because the phenomenon occurs on almost all hill slopes, 
including the afforested ones. Superficial and shallow landslides 
superposes on large, old and deep landslide bodies (Ielenicz 
et.al., 1999; Armas et. al. 2004). 
Landslide spatial occurrence in Romania was the subject of 
some recent papers. Most of the authors focuses on landslide 
susceptibility issues (Armas et al, 2004; Chitu et al., 2009; 
Mihai et al., 2010 etc.) and some of them focused also on 
hazard analysis (Sandric, 2008 unpublished) at different scale 
approaches, from local and regional analysis and to the whole 
Romanian territory (Balteanu et. al., 2010 in press). The GIS 
modeling (deterministic and probabilistic approaches) and 
mapping techniques are largely used within these papers 
The aim of the paper is to develop a semi-automated algorithm 
for landslide body’s identification and classification. The 
algorithm is based on object based image analysis 
(segmentation and classification) applied on high resolution air- 
photos in combination with lithology, morphometrical features 
and meteorological data 
For the areas with complex morphodynamic features like the 
Subcarpathians, the image visual interpretation and field 
surveys are usually a time consuming step (Sandric, 2009, 
Moine, 2009, Tapas, 2010). It is usually very difficult to extract 
all the landslide bodies even on smaller areas (a catchment or a 
small administrative unit like a town or a community), because 
they have a complex typology (landslides, earthflows etc), they 
occurs on the whole slope area, the landslide bodies of different 
generation are superposed 
1.2 Main contributions in the international literature 
International contributions on landform automatic classification 
are quite recent approaches (Dragut, Blaschke, 2006). Landslide 
analysis on satellite imagery is an older research field 
(Mantovani et al., 1996, McDermid, Franklin, 1994, Brardinoni 
et al., 2003; van Westen, 2003; etc.). Since 2006, there is an 
increasing interest in developing algorithms for landslide 
classification, based on spectral data (Borghuis et al., 2007) as 
well as on digital elevation data (van van Asselen, 
Seijmonsbergen, 2006). Landslide mapping based on remote 
sensing imagery object based image analysis and classification 
(Nichol et al., 2005; Moine et al., 2009), remote sensing 
Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.