EXTRACTION OF WIND EROSION OBSTACLES 
BY INTEGRATING GIS-DATA AND STEREO IMAGES 
Y. L Zhang ** 
b 
" School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, Hubei, 430079, P.R. China 
Supresoft Inc, 3-2#Building, Guandong Science Park, 2"! Guanshan Road, Wuhan, Hubei, 430074, P.R. China 
yjzhang@supresoft.com.cn 
Theme Session 12 
KEY WORDS: Extraction, Integration, GIS, Image, Analysis, Vegetation, Classification 
ABSTRACT: 
Data integration is a very important strategy to obtain optimum solutions in geo-scientific analysis, 3D scene modelling and visualization. 
This paper mainly focuses on the integration of GIS-data, stereo aerial imagery and DSM to derive automatically wind erosion obstacles 
in the open landscape to enhance the Digital Soil Science Map of Lower Saxony in Germany. The extracted wind erosion obstacles can be 
used to derive wind erosion risk fields for soil monitoring and preserv 
ation. GIS-data is used as prior information for the object extraction. 
The GIS-objects roads, field boundaries, rivers and railways from GIS database can represent initial search areas for extracting wind 
erosion obstacles, which are often located parallel and near to them. Wind erosion obstacles are divided in the semantic model into hedges 
and tree rows, because of different available information from the GIS-data, although their extraction strategies are similar. Different 
approaches, such as segmentation by NDVI and CIE L*a*b, edge e 
xtraction, linking, grouping and verifying with 3D information, are 
e eo g 
combined to extract the objects of interest. The extracted wind erosion obstacles are integrated into a semantic model, described by their 
3D appearance in geometry, together with 2D elongated shadow regions in a known direction according to 3D information and sunshine. 
I. INTRODUCTION 
In geo-scientific analysis, estimation of potential erosion areas, 
soil monitoring and preservation, the integration of different data 
sources is an important strategy to obtain overall solutions. To 
facilitate automated object extraction from aerial imagery, the use 
of prior knowledge is essential (Baltsavias 2002, Straub 2003). 
This paper mainly focuses on extraction of wind erosion obstacles 
by data integration to enhance the Digital Soil Science Map of 
Lower Saxony (Germany) in the open landscape. GIS-data, aerial 
stereo imagery and digital surface models (DSM) are used as 
sources of information. The extracted wind erosion obstacles can 
be used to derive potential wind erosion risk fields together with 
information of prevailing wind direction, field width and soil 
parameters. 
Digital color infrared (CIR) imagery is an important development 
in data acquisition and update especially for vegetations (Englisch 
and Heipke 1998). The Normalized Difference Vegetation Index 
(NDVI) is widely used in photogrammetry and remote sensing 
applications, such as monitoring of vegetation condition and 
production in different environmental situations (Prince 1991), 
change detection (Lyon et al 1998), extracting of trees in urban 
areas (Straub 2003) and detecting smoke effects of vegetations 
(De Moura 2003). The CIE L*a*b color space is mainly used in 
computer vision communities for image analysis and industrial 
applications (Pierce 1994, Campadelli 2000, Lebrun 2000). 
Nevertheless, it seems not of much interest by photogrammetrists 
despite its powerfulness in image classification and analysis. 
Although considerable results have been achieved, the extraction 
of vegetation objects from high-resolution imagery is still not in 
an advanced period (c.f. Heipke et al. 2000, Straub 2003). Prior 
work of extraction of wind erosion obstacles in the open 
landscape is much more marginal. Our specific task is to extract 
wind erosion obstacles to enhance the soil map for potential 
utilization of soil monitoring and preservation. The related work 
of the specific task, such as semantic modeling, extraction and 
updating of field boundaries (Butenuth 2003) are not of interest in 
this paper, although wind erosion obstacles are often field 
boundaries or at least can help to extract field boundaries. 
General strategy and the data sources used for extraction of wind 
erosion obstacles are described in the next section. Afterwards, 
the detailed approach for extraction of wind erosion obstacles, 
including image segmentation with NDVI and CIE L*a*b, line 
extraction and grouping, verifying with 3D information will be 
presented. Experimental results of automatically derived wind 
erosion obstacles in the open landscape are given in section 4 to 
demonstrate the potential of the proposed approach. Finally, 
conclusions are given and further work is highlighted. 
2. GENERAL STRATEGY AND DATA SOURCES 
2.1 General strategy 
The enhance of Digital Soil Science Map with extracted wind 
erosion obstacles by integrating GIS-data, CIR stereo imagery as 
well as DSM are the content of the current work. GIS-data 
represents an initial scene description of the open landscape of 
interest. The object wind erosion obstacle is divided in hedge and 
tree row, because there are different information available from 
GIS data, although their extraction strategies are similar. The 
nc 
ex 
lit 
lar