Back to course web page or main web page

CE691M - Geomatics Engineering Seminar, Spring 2008

Time: 1:30 – 2:20pm, Wednesdays; Location: CIVL 3153 (if not otherwise indicated)

April 23, 2008, at Stewart 209

Bert Meeus, University of Leuven, Belgium

Evaluating the possibilities of TerraSAR-X, with emphasis to interferometry and object oriented land use classification

TerraSAR-X is a new very high resolution imaging radar satellite, launched in June 2007. Possible applications of the TerraSAR-X imagery are investigated in this study. The emphasis of the study is laid on 2 applications. The interferometric capabilities of the TerraSAR-X sensor are investigated for dike monitoring purposes of the dikes of the river Scheldt, close to the village of Temse, Belgium. The second application is a land use classification of a test area east of the Belgian city of Antwerpen. For this application, TerraSAR-X images were used together with IKONOS data and a TeleAtlas GIS vector layer.

Bert Meeus graduated from the KULeuven (Catholic University of Leuven, Belgium) in 2006 as a bio-engineer in land and forest management. He is now enrolled in the Advanced Master program of Earth Observation, a program set up jointly by K.U.Leuven and Purdue University. For the program, Bert stayed one semester at Purdue University. He is doing his master thesis in cooperation with GIM, a Belgian company specialized in GIS and remote sensing. Promoter of the thesis is professor Gerard Govers.

April 16, 2008, at Stewart 209

An Van Delm, University of Leuven, Belgium

Green in the expanding urban and semi-urban complex: application of detailed field data and IKONOS imagery

Urbanization of the countryside may occur according to different patterns and densities and can eventually lead to stabilizing complexes composed of new and former landscape elements. Unlike the general perception, and especially in the morphologically looser patterns of urbanization, the outcome may be more rather than less greenery in the landscape. In order to help investigate at regional level the spatial and functional relationships of hard elements of urbanization (roads, buildings etc.) with green elements, more specifically woody vegetation, we call in very high resolution IKONOS data, to be correlated with a contemporary field survey in the vicinity of Roeselare in western Belgium.

The exploratory research investigates the use of this very high-resolution imagery for the quantification of urban green variables (area, height and volume). The method would like to overcome the gap between the popularity of the images, however the difficulties that still exist by processing it. Therefore the quantification of the urban green variables is investigated by defining the qualities of the multispectral information, namely by vegetation indices. First, the investigation defines the index, which extracts the (semi-) urban vegetation most accurately. After this the selected vegetation index will be used for the quantification of the present structural vegetation elements. The possibilities of this easy to use method will be investigated, as well as compared to an existing object-based classification performed by the Flemish Geographical Information Agency (FGIA). The results can also be used to upscale to broader regional ranges in using coarser resolution satellite imagery. This whole research helps to assess how urbanization rather than just be an unsustainable process or equivalent to desertification, can also partly be a process contributing to enhanced sustainability conditions through its associated vegetation. 

Brief bio:

July 2006 Graduated as Master in Engineering-Architecture, K.U. Leuven,

2006 - now: Student in the (advanced) master program of Earth Observation, a program set up jointly by K.U.Leuven and Purdue University. For the program, An stayed one semester at Purdue University. Promoter of the research project is professor Hubert Gulinck.

April 9, 2008, at Stewart 209

Jasper Van doninck, University of Leuven, Belgium

Agricultural Area Estimation from Medium Resolution Imagery in Ethiopia - Influence and Integration of Elevation Data

A methodology is presented for the estimation of agricultural area statistics from MERIS fAPAR time series using limited ground reference data. Early crop area estimates at national and local scale can be used for supporting food policies. In a first step, high resolution images (SPOT HRVIR) over selected representative sites in the northern part of Ethiopia are classified using ground reference data obtained from transect sampling. The produced hard classification is than degraded to produce medium resolution area fraction images, which are in turn used as reference for the soft classification of the medium resolution data. An artificial neural network is used for both high and low resolution classification. Elevation in Ethiopia ranges from below sea-level to over 4000 m and affects classification accuracies and procedures by the creation of different climate zones, relief displacement, High resolution (90m) elevation data, obtained by the Shuttle Radar Topography Mission (SRTM), is nowadays available worldwide and integrated in this study in the preprocessing, high resolution and low resolution classification stages. The produced results are compared with official statistics.

Jasper Van doninck obtained a master degree in geography from the Free University of Brussels in 2006. He is now enrolled in the Advanced Master of Earth Observation, a program set up jointly by K.U.Leuven and Purdue University. For the program, Jasper stayed one semester at Purdue University. He is doing his master thesis in cooperation with VITO, the Flemish (public) institute for technological research. Promoter of the research project is Professor Jos Van Orshoven.

April 2

Prof. Darrell G. Schulze, Agronomy Department

with contribution from Phillip R. Owens, and George E. Van Scoyoc

Taking GIS to the Field

Soils occur in the field in predictable, repeating patterns, the result of five major soil forming factors that can vary over length scales ranging from a few meters to tens or hundreds of kilometers. Teaching students how to recognize spatial differences in soils and landscapes, and how to draw inferences from them, is a significant instructional challenge. The right visual (a particular map or sequence of maps), presented at the right time (while students are in the field) makes complex relationships clear and easy to remember. Paper maps, however, have major limitations. For the past 3 years, we have been using rugged Tablet PCs equipped with GPS receivers and GIS software to display maps of our study areas while we are in the field. Students have responded enthusiastically and learning is enhanced significantly. As instructors, we have learned new things that were not apparent to us before, and we are now able to teach these subtle and complex relationships more effectively. In addition to the pedagogic aspects, we’ll discuss the mechanics of taking up to 14 Tablet PCs to the field on an almost weekly basis. Several tablet PCs and accessories will be available for hands-on examination.

March 26

Prof. William Emery, University of Colorado

Spatial satellite remote sensing; past, present and future

We review satellite remote sensing from the early days of meteorological sensors up to the present series of land remote sensing satellite. We end by reviewing some work with very current methods for analyzing the highest spatial resolution images from QuickBird and WorldView 1 satellites.

BSc in Mechanical Engineering BYU 1971

PhD Physical Oceanography U. of Hawaii 1975

Professor of Aerospace Engineering Sciences, Univ of Colorado 1987 - present.

Adjunct professor of Informatics Engineering, Tor Vergata Univ, Rome, Italy 2006 - present.

Involved with land remote sensing studies since 1986. Present of Agrisat 1987 - 1997; forecast crop production from satellite data. Involved in high resolution satellite image analysis since 2000.

March 18, 19, 20 (Tuesday, Wednesday, Thursday)

Prof. Dr.-Ing. Wolfgang Forstner

Institute of Photogrammetry, University of Bonn, Germany

Notice the different locations on different days.

Lecture 1: Tuesday 1:30-2:45 in TV-studio POTR 268

Detecting and Reconstructing Buildings from Aerial Images and LIDAR Data. 3D-city models are of broad interest for planning, tourism, etc. as can be seen from the increasing availability in Google-map. This lecture gives an overview on the strategies and methods used for automatic building detection and reconstruction on large images.

Lecture 2: Wednesday 1:30-2:20 CIVL 3153

Automatic image orientation. Orientation procedures are the basic requisite for any photogrammetric 3D-evaluation of the images. The lecture presents methods for automatic relative orientation of multiple images without targeted points and discusses the matching and reconstruction strategy.

Lecture 3: Thursday 3:00-4:15 in TV-studio POTR 268

Uncertain Geometric Reasoning for the Reconstruction of Polyhedral Objects. Projective Geometry is a versatile tool not only for representing the camera geometry but also for reconstructing man-made objects. The inherent uncertainty in the given measurements suggests to fuse projective geometry with statistics to exploit the advantages of both. The lecture gives an introduction into this fusion and demonstrates its versatility in object reconstruction.

February 27

Prof. Daniel Aliaga, Department of Computer Science

Content-Aware Urban Layout Editing

with contribution from Carlos Vanegas (PhD Student, CS), Bedrich Benes (Asst Prof, CGT)

In this talk, we present an interactive system for efficient and intuitive urban layout creation and modification. Our key inspiration comes from recent content-aware image editing methods. These techniques use high-level image information to enable fast and intuitive operations such as “smart” resizing and image retargeting beyond that provided by naïve image processing. In a similar fashion, we use the information provided by aerial photographs and GIS or by procedural methods to generate an urban layout and provide the tools and algorithms for the efficient and interactive creation of new urban spaces and the modification and extension of existing urban areas. The end result is the intuitive, rapid, and interactive generation of new or edited urban layouts which include visual imagery as well as a valid and feasible underlying structure including a hierarchy of streets and intersections, and parcel information. We demonstrate our system by editing and extending several existing cities, ranging from hundreds to thousands of city blocks and parcels and by quickly generating new cities in the style of other cities.

February 20

Amélie Davis, Dept. of Forestry and Natural Resources

Title: Estimating Parking Lot Coverage in the Great Lakes Region

We have determined a method to estimate the areal footprint of parking lots in the Great Lakes Basin as a first step toward creating a basis for enhanced parking lot regulation for decision makers at the level of the individual counties but also applicable to cities or states.  Urban sprawl combined with population growth have been the main agents responsible for this era of paved landscapes.  Parking lots are the insidious partner of human built structures and often are larger than building footprints themselves, yet parking lots are virtually everywhere and largely unregulated.  In this study we determine the areal footprint of parking lots in the Great Lakes Basin and outline policies to provide smart growth of parking lots.  Our presentation addresses the question: "what is the areal footprint and the ensuing economic and ecological consequences of parking lots?". To that effect, we use high resolution aerial photography and GIS to estimate the areal footprint of parking lots in relationship to: (1) the total urban area; (2) the number of parking spaces versus the population of the county; and (3) the distribution of parking spaces by land use category for a high density urban cover portion of the study area. Finally, we develop a set of metrics useful to relate our findings to urban planners and county regulators and 4) we outline the necessary steps to extrapolate our findings to the United States scale and from that draw conclusions as to a possible sustainable path to controlled growth.

February 6

Wonkook Kim, Geomatics Department. School of Civil Engineering, Purdue University

Title : Multi-resolution Manifold Learning for Classification of Hyperspectral Data

Nonlinear manifold learning algorithms assume that the original high dimensional data actually lies on a low dimensional manifold defined by local geometric distances between samples. However, most of the traditional methods have focused only on the spectral distances in calculating the local dissimilarity of samples whereas in case of image-like data the disposition of the image samples would provide useful information. As a framework of integrating spatial and spectral information in terms of image samples, the hierarchical segmentation method (RHSEG) is introduced and the manifold structure based on the segmentation results will be investigated. A manifold update scheme is also proposed where the spatial information is transferred through the segmentation hierarchy.

January 30

Prof. Guofan Shao, Department of Forestry and Natural Resources, Purdue University

Classification Accuracy: User’s Perspectives

Remote sensing technology has advanced markedly during the past decades.  Remotely sensed data with various resolutions are available and image processing programs are easy to use. Developing map products with remote sensing data is a convenient task. However, little attention has been given to the effects of classification accuracy on the applications of map products. Thematic maps derived from image classification are not always the final product from the user’s perspective. Because all image processing or classification inevitably introduces errors into the resultant thematic maps, any subsequent quantitative analyses will reflect these errors. Unfortunately, the error propagation processes have nonlinear relationships with classification accuracy.  For example, a small difference in classification accuracy resulted in a large difference in landscape index values when thematic maps are used in quantify landscape patterns. It is important for analysts to make every effort to increase classification accuracy; it is important for users to be aware of misclassification errors and their implications.

January 23

Darion S. Grant, Geomatics Department. School of Civil Engineering, Purdue University

LIDAR Strip adjustment with TIN matching

LIght Detection And Ranging (LIDAR) provides a dense sampling and cost effective tool for the acquisition of discrete elevation data. The dataset is not error-free and often requires some adjustment procedure to reduce or remove the effects of any unmodeled systematic errors. Many approaches organize the data into a Triangular Irregular Network (TIN) structure and perform linear interpolation of the triangular faces to extract surface information for the adjustment. This presentation proposes an alternative approach to the interpolation procedure when the dataset represents the Earth’s terrain and involves the use of the Thin Plate Spline (TPS) function. Preliminary experiments with synthetic data showed that the proposed approach outperformed the planar interpolation approach by about 35% in cross validation and validation tests and by about 30% in 3D translation and 3D similarity strip adjustment procedures.

January 16

Chad M. Schaeding, Geomatics Department. School of Civil Engineering, Purdue University

Geostatistical Spatial Data Fusion

This period of history is an exciting time for the spatial sciences. More diverse data collection instruments are gathering, even more information about our environment. These developments tax our ability to handle and process this burgeoning amount of data. Models are becoming more sophisticated because our understanding of the relationships between phenomena is getting better. Once processed, we are now faced with the daunting task of fusing newly collected data with legacy information. We will describe the geostatistical paradigm for spatial data conflation. This paper will explore the principles of the geostatistical tools and how that can be applied in practice. Two case studies are investigated to show some of the benefits and limitations to the presented tools.

Tentative speakers for the rest of the semester

Davis, Amelie Y, Department of Forestry and Natural Resources

Qianlai Zhuang,  Assistant Professor, Department of Earth & Atmospheric Sciences

Darrell G. Schulze, Professor, Department of Agronomy

Wofgang Forstner, Professor, University of Bonn, Germany

Daniel G. Aliaga, Department of Computer Science