of the Agency and no official endorsement should be inferred.

A classification tree based approach for building detection was tested. A digital surface model (DSM) derived from last pulse laser scanner data was first segmented and the segments were classified into classes ‘ground ’ and ‘building or tree ’ on the basis of preclassified laser points. ‘Building and tree ’ segments were further classified into buildings and trees by using the classification tree method. Four classification tests were carried out using different combinations of 44 input attributes. The attributes were derived from the last pulse DSM, first pulse DSM and an aerial colour ortho image. In addition, shape attributes calculated for the segments were used. The attributes of training segments were presented as input data for the classification tree method, which constructed automatically a classification tree for each test. The trees were then applied to classification of a separate test area. Compared with a building map, a mean accuracy of almost 90 % was achieved for buildings in each test. The classification tree method appeared to be a feasible and highly automatic approach for distinguishing buildings from trees. If new data sources become available in the future, they can be easily included in the classification process. The results also suggest that satisfactory building detection results can be obtained with different combinations of input data sources. By using a statistical method, it is possible to find useful attributes and classification rules in different cases. The use of an aerial image or both first pulse and last pulse laser scanner data does not necessarily improve the results significantly, compared with a classification that uses only last pulse laser scanner data. 1.

Cooperative Agreement Number CR-825506. The contents have not been subjected to Agency review and therefore do not necessarily reflect the views of the Agency and no official endorsement should be inferred.

Number CR-825506. The contents have not been subjected to Agency review and therefore do not necessarily reflect the views of the Agency and no official endorsement should be inferred.

We consider the problem of correcting misclassifications in images by using context based or spatial information. We describe a graph-based method for correcting misclassifications that occur in primary local image recognition. The proposed method is applied in a training-based optimization framework, using genetic algorithms. Numerical simulation results are presented to confirm that once the optimal parameters are obtained by training on one image, satisfactory performance is obtained by using the optimal operator on another significantly different image.

on the authority of the Rector Magnificus of Wageningen University,

In order to understand causes and effects of disturbance and fragmentation on flora and fauna, a time series on land cover change is needed as basis for the BIOTA-East Africa project partners working in western Kenya. For 7 time steps over the past 30 years Landsat data were collected for Kakamega Forest and its associated forest areas. Preprocessing involved georeferencing and radiometric corrections. In a first step the time series is evaluated via a threshold analysis distinguishing between “forest ” and “non-forest”. Even though a temporally changing pattern of forest losses and replanting is observed, in total no major change in forest-covered area is revealed. Therefore, a supervised multispectral classification is performed distinguishing between classes at the ecosystem level. Ground truthing for the historical imagery is done with the help of maps showing vegetation types or land cover. Actual land cover verification is based on amateur photographs taken from an aeroplane as well as on terrain references. For classification the maximum-likelihood decision rule is applied considering bands 3, 4, 5, 7 plus 7/2 for TM/ETM+ imagery and 1, 2, 3 and 4 for MSS-data, repectively. If available, scenes from both the rainy and dry seasons are made use of. From planned 17 land cover classes 12 can be realized, of which 6 belong to forest formations. A shortcome is that plantation forest of Maesopsis eminii (planted mixed in with other indigenous tree species) cannot be separated. Nevertheless, the classification results form a solid basis for a consistent and detailed evaluation of forest history between 1972 and 2001. Analyses presented include graphs of change in land cover class areas over time as well as such allowing for true change detection with transitions between the different classes. 1.

A Geographic Information System (GIS) viewshed is the result of a function that determines, given a terrain model, which areas on a map can be seen from a given point(s), line or area. In the communications industry, this function has been used to model radio wave coverages and to site transceiver towers for cellular phones. However, there are errors involved with this function and, without the requisite data, it cannot account for building heights that may affect visibility in urban areas. This paper examines the ability to accurately show line-of-sight (LOS) radio wave coverages in order to establish the viability of replacing existing field methods with GIS viewshed analysis. An origin point capable of supporting a line-of-sight radio wave transmitter was chosen from within the Virginia Tech campus study area. A viewshed analysis was performed with ESRI's ArcView GIS, using this site as the observation point and a 30-meter resolution Digital Elevation Model (DEM) from the US Geological Survey. To check the accuracy of the viewshed, we transmitted at 27.5 GHz, a LOS frequency that has properties common in the wireless telecommunications industry. We

Digital Elevation Models (DEMs) are computer-generated representations of the earth’s surface. These surfaces can be used to predicted Line-of-Sight (LOS) radio propagation. DEM resolution can affect the results of this prediction. This study examines the effect of DEM resolution on accuracy by comparing varied resolution terrain data for a portion of Blacksburg, Virginia using the prediction of ESRI’s ArcView ® viewshed algorithm. Results show that resolutions between one-meter and thirty-meters have little effect on the aggregate accuracy of the viewshed. iii Acknowledgements I would like to express my most sincere gratitude to Dr. Laurence (Bill) Carstensen for his guidance and ultimate patience and support throughout this study. In addition, I would like to thank Dr. Dennis Sweeney for his help and patience in teaching me the fundamentals of RF needed for this study. He was also essential in providing equipment that was his personal equipment, and borrowed equipment worth more than the cost of my entire college education. I would also like to thank my other two committee members, Dr. Charles Bostian, and Dr. James Campbell for their support. I would like to thank the Center for Wireless Telecommunication for their support throughout the study. I would like to my family for their support throughout, and especially my fiancée Amy who has loved and encouraged me from beginning to end. iv

Fugitive (or airborne) dust is a primary cause of decreased air quality, as well as being a potential health hazard. Urban and agricultural areas are of particular interest as fugitive dust sources because of their potential for releases during soil disturbance, ongoing industrial and commercial processes, and agricultural activities. Typical strategies for assessing and monitoring fugitive dust source areas include numerical modeling of atmospheric circulation patterns, field assessments, and collection of dust samples using various methods. Analysis of remotely sensed multi-spectral data provides another alternative for identifying fugitive dust source, transport, and sink areas. Multi-spectral (visible to shortwave infrared) data acquired by the Enhanced Thematic Mapper Plus (ETM+) instrument on board the Landsat 7 satellite is used to perform land-cover classifications for the Nogales, AZ, region. Data acquired during the winter of 2000 and the summer of 2001 are used to assess seasonal variations and detect land-cover changes of significance to dusttransport processes. An expert system approach using spectral, textural, and vegetation abundance data is used to classify the ETM+ data into land-cover types important to dust-transport models. The determined overall accuracy of the land-cover classifications is 74 percent. These results can be used to identify (and calculate areal percentages of) fugitive dust source, transport, and sink regions. This spatially explicit, digital data product is useful both as an input into dust-transport models and as a check on the results of such models.