In: Wagner W„ Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
275 
AUTOMATIC QUALTIY CONTROL OF CROPLAND AND GRASLAND GIS OBJECTS 
USING IKONOS SATELLITE IMAGERY 
P. Helmholz*, C. Heipke, F. Rottensteiner 
IPI - Institute of Photogrammetry and Geolnformation, Leibniz University Hannover, Nienburger Str. 1, 30167 
Hannover, Germany - (helmholz, heipke, rottensteiner) @ipi.uni-hannover.de 
Commission VII 
KEY WORDS: Automation, GIS, Quality Control, Verification, Updating 
ABSTRACT: 
As a consequence of the wide-spread application of digital geo-data in Geoinformation Systems (GIS), quality control has become 
increasingly important. A high degree of automation is required in order to make quality control efficient enough for practical 
application. In order to achieve this goal we have designed and implemented a semi-automatic technique for the verification of 
cropland and grassland GIS objects using 1 m pan-sharpened multispectral IKONOS imagery. The approach compares the GIS 
objects and compares them with data derived from high resolution remote sensing imagery using image analysis techniques. 
Textural, structural, and spectral features are assessed in a classification based on Support Vector Machines (SVM) in order to check 
whether a cropland or grassland object in the GIS is correct or not. The approach is explained in detail, and an evaluation is 
presented using reference data. Both the potential and the limitations of the system are discussed. 
1. INTRODUCTION 
Today, many public and private decisions rely on geospatial 
information. Geospatial data are stored and managed in 
Geoinformation Systems (GIS) such as the Authoritative 
Topographic Cartographic Information System (ATKIS) or the 
Digital Landscape Model (DLM-DE) in Germany (Arnold, 
2009). In order for a GIS to be generally accepted, the 
underlying data need to be consistent and up-to-date. As a 
consequence, quality control has become increasingly 
important. In the European Norm DIN EN ISO 8402 (1995), 
quality is defined as the “Totality of characteristics of an entity 
that bear on its ability to satisfy stated and implied needs”. In 
the context of GIS this means that the data model must 
represent the real world with sufficient detail and without any 
contradictions (quality of the model). Secondly, the data must 
conform to their specification (quality of the data). There are 
four important measures for quality control of geodata: 
consistency, completeness, correctness, and accuracy (Joos, 
2000). Only the consistency can be checked without any 
comparison of the data to the real world. All the other quality 
measures can be derived by comparing the GIS data to the real 
world, as it is represented in aerial or satellite images. In order 
to reduce the amount of manual work required for quality 
control, a high degree of automation is required. In this paper, 
we describe a method for the verification of agricultural objects 
for quality control that is based on 1 m pan-sharpened 
multispectral IKONOS images. The focus will be on the 
separation of grassland and cropland objects for the quality 
management of ATKIS, because it has been found that these 
classes are not easily separated, e.g. (Regners & Prinz, 2009). 
After giving an overview on related work in Section 2, our new 
approach is described in Section 3. First results are presented in 
Section 4. The paper concludes with a discussion about the 
potential and the limitations of this approach. 
2. RELATED WORK 
Lu and Wenig (2007) gave an overview about the state of the 
art classification techniques. They emphasise that besides 
textural and spectral approaches, approaches using context 
information (such as structures) become more important with 
increasing resolution of the images. In this section we briefly 
review approaches for extracting different agricultural object 
types based on textural, structural and spectral features using 
high resolution images. 
Textural features are related to local spatial patterns of grey 
levels inside an object. There have been quite a few attempts to 
use the textural characteristics for the classification of different 
agricultural object classes. For instance, autocorrelation is used 
by Warner and Steinmaus (2005) to identify orchards and 
vineyards in IKONOS panchromatic imagery. After defining a 
square kernel and after radiometric normalization, the 
autocorrelation is determined for the cardinal directions and 
both diagonals, which results in one autocorrelogram per 
direction. An orchard pixel is detected if an orchard pattern is 
identified in more than one autocorrelogram centred on that 
pixel. This method assumes the rows of plants to be equally 
spaced. Rengers and Prinz (2009) use the neighbourhood grey- 
tone difference matrix (NGTDM) to classify cropland, forest, 
water, grassland and urban areas in aerial and IKONOS images. 
This method is based on the differences of the grey values of 
two pixels and the differences of the grey values of the local 
neighbours, from which textural features such as coarseness, 
complexity and textural strength are derived. The results 
presented in (Rengers & Prinz, 2009) show that with the 
exception of grassland and cropland the classes mentioned 
above can be distinguished well. A similar conclusion is drawn 
by Busch et al. (2004j, who apply a texture-based classification 
method based on Markov random fields (Gimel’farb, 1996) to 
aerial and IKONOS satellite images. Their method is well- 
suited to classify settlement areas, industrial areas, forests, and 
* Corresponding author