10

Object-Based Image Analysis for

Thomas Blaschke, Kasper Johansen, and Dirk Tiede

Contents
10.1 Introduction......................................................................................................................... 241
10.2 Object-Based Image Analysis for Remotely Sensed Vegetation Mapping................. 243
10.2.1 Object-Based Image Analysis................................................................................ 243
10.2.2 Object-Based Image Analysis and Increasing Spatial Resolution................... 244
10.2.3 Remote Sensing of Vegetation............................................................................... 246
10.2.4 State-of-the-Art Object-Based Image Analysis of Vegetation........................... 248
10.2.4.1 Vegetation Inventory and Classification............................................... 248
10.2.4.2 Change Detection..................................................................................... 248
10.2.4.3 High Spatial Resolution Optical Data................................................... 249
10.2.4.4 Light Detection and Ranging Data........................................................ 249
10.2.4.5 Incorporating Nonspectral Information............................................... 250
10.3 Case Studies of Object-Based Image Analysis for Mapping Vegetation.................... 251
10.3.1 Case Study 1: Mapping Plant Projective Cover within Riparian Zones
Using Object-Based Image Analysis and Light Detection and Ranging Data��� 251
10.3.1.1 Introduction.............................................................................................. 251
10.3.1.2 Methods..................................................................................................... 251
10.3.1.3 Results and Discussion........................................................................... 255
10.3.1.4 Summary................................................................................................... 257
10.3.2 Case Study 2: Tree Crown Extraction in a Low Mountain Range Area
from UltraCamX-Derived Surface Models Using Object-Based Image
Analysis and Grid Computing Techniques������������������������������������������������������� 257
10.3.2.1 Introduction.............................................................................................. 258
10.3.2.2 Methods..................................................................................................... 258
10.3.2.3 Results........................................................................................................ 261
10.3.2.4 Discussion and Outlook.......................................................................... 263
10.4 Summary and Discussion................................................................................................. 263
10.5 Future Research.................................................................................................................. 265
References...................................................................................................................................... 266

10.1  Introduction
Environmental monitoring requirements, conservation goals, spatial planning enforce-
ment, and ecosystem-oriented natural resources management, to name just a few ­drivers,
lend considerable urgency to the development of operational solutions that can extract

241
242 Advances in Environmental Remote Sensing

tangible information from remote sensing data. The “workhorses” of satellite data
generation, such as the Landsat and System Pour l’Observation de la Terre (SPOT) satel-
lites or the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)
and Moderate Resolution Imaging Spectroradiometer (MODIS) instruments, have become
important in global and regional studies of biodiversity, nature conservation, food secu-
rity, deforestation impact, desertification monitoring, and other application fields. With
the increasing spatial resolution of the “1-m generation” of IKONOS (launched in 1999),
QuickBird (2001), and OrbView (2003) sensors and the increasing availability of airborne
optical digital high spatial resolution imaging sensors and laser scanners, new application
fields that had previously been the domain of analog airborne remote sensing can now
be tackled with increasing flexibility and precision, at reduced costs and for remote areas
not previously accessible. In late 2007, the first commercial satellite with a spatial resolu-
tion of less than half a meter (WorldView-1; 0.44-m panchromatic) became operational,
followed by multispectral sensors such as GeoEye-1 (2008) and WorldView-2 (2009), pro-
viding pan-sharpened subhalf-meter resolution image data. At present, we see security
applications, vehicle detection, urban mapping, and vegetation assessment applications
developing rapidly, in terms of both number and sophistication. By simplification and
generalization, we can distinguish two major trends: (1) an increasing amount of data is
being produced in an ever-broadening range of spatial, spectral, radiometric, and tem-
poral resolutions, including the aforementioned high spatial resolutions; and (2) national
and supranational programs and systems are being developed for regular or on-demand
vegetation surveys.
Suitable remotely sensed image data for mapping vegetation cover and properties at
different spatial scales (from global to local) are becoming increasingly available. There
is an extensive body of work covering the usefulness of these image data and the associ-
ated methods for vegetation mapping and monitoring (Coppin et al. 2004; Lu et al. 2004).
However, mapping and monitoring of biophysical vegetation parameters and extraction
of stand parameters such as height, age, and foliage projective cover on a regular and
cost-effective basis had not been possible until recently due to a lack of data with suf-
ficient spatial resolution (Congalton et al. 2002; Gergel et al. 2007; Kayitakire, Hamel,
and Defourny 2006). The availability of data from high spatial resolution sensors such
as the IKONOS, QuickBird, GeoEye-1, and WorldView-2 satellite sensors and airborne
multispectral, hyperspectral, and light detection and ranging (lidar) sensors has opened
up new opportunities for the development of operational mapping and monitoring of
small features such as individual tree crowns and narrow riparian zones (Hurtt et al.
2003). The capacity to map these small features and related vegetation structural param-
eters has improved over the last decade through the use of object-based image analysis
(OBIA).
This chapter assesses the potential of OBIA for vegetation mapping and monitoring.
The acronyms OBIA and GEOBIA, which stands for geospatial OBIA, are both used herein
interchangeably. The chapter summarizes trends in vegetation remote sensing and reflects
very briefly on underlying concepts such as image segmentation, which is much older
than the popularized commercial software of today. The goal of this chapter is to provide
a review of OBIA applications for remote sensing of vegetation. Sections 10.2.1 through
10.2.4 provide a background on the use of OBIA for remote sensing of vegetation and infor-
mation on how the main remote sensing systems used for vegetation/forestry applications
differ from one another. The intent is not to review the sensors, but to focus on OBIA
methods used for vegetation studies, which are also demonstrated through two separate
case studies in this chapter.