517 Detection and Vectorization of Roads from Lidar Data
Simon Clode, Franz Rottensteiner, Peter Kootsookos, and Emanuel Zelniker
A method for the automatic detection and vectorization of roads from lidar data is presented. To extract roads from a lidar point cloud, a hierarchical classification technique is used to classify the lidar points progressively into road and non-road points. During the classification process, both intensity and height values are initially used. Due to the homogeneous and consistent nature of roads, a local point density is introduced to finalize the classification. The resultant binary classification is then vectorized by convolving a complex-valued disk named the Phase Coded Disk (PCD) with the image to provide three separate pieces of information about the road. The centerline and width of the road are obtained from the resultant magnitude image while the direction is determined from the corresponding phase image, thus completing the vectorized road model. All algorithms used are described and applied to two urban test sites. Completeness values of 0.88 and 0.79 and correctness values of 0.67 and 0.80 were achieved for the classification phase of the process. The vectorization of the classified results yielded RMS values of 1.56 m and 1.66 m, completeness values of 0.84 and 0.81 and correctness values of 0.75 and 0.80 for two different data sets.
537 A Survey on the Need for Airborne Lidar Training
Chris Hopkinson, Sorin Popescu, Martin Flood, and Robert Maher
Two questionnaires were sent out to over 600 members of the international lidar academic research and commercial mapping community, and a lidar research and training workshop was hosted in Halifax, Canada. The purpose of the questionnaires and the workshop was to better understand the status of, and needs for training within the lidar community. The results demonstrate that there is a clear need for training within both the end user and service provider sectors of the professional lidar community. It is speculated that although specific training needs differ, in terms of volume the end user community’s need is at least an order of magnitude greater than in the service provider sector. Regarding training priorities, there appears to be some clear stratification between the needs of end users and service providers. In general, practical experience and“hands on” training methods were considered more useful for those entering into lidar related employment, but this perception was not shared by academics. Also, results indicated that “end user applications” were the priority topic in the end user academic and government communities, while in the lidar industry, training priorities were related to more technical and operational topics such as“data processing” and “project management.” Within the lidar project workflow, six areas of responsibility were identified within the end user and service provider sectors (service provider operators, data processors, and project managers; and end user clients, project managers, and data processors), each of which having different training needs. The results of this study are being used by the Applied Geomatics Research Group to develop a suite of lidar training curricula from workshop seminars to industry-sponsored project-based internship programs.
547 A Filtering Strategy for Interest Point Detecting to Improve
Repeatability and Information Content
Qing Zhu, Bo Wu, and Neng Wan
This paper compares several stereo image interest point detectors with respect to their repeatability and information content through experimental analysis. The Harris-Laplace detector gives better results than other detectors in areas of good texture; however, in areas of poor texture, the Harris-Laplace detector may be not the best choice. A feature-related filtering strategy is designed for the Harris-Laplace detector (as well as the standard Harris detector) to improve the repeatability and information content for imagery with both good and poor texture: (a) the local information entropy is computed to describe the local feature of the image; and (b) the redundant interest points are filtered according to the interest strength and the local information entropy. After the filtering process, the repeatability and information content of the final interest points are improved, and the mismatching then can be reduced. This conclusion is supported by experimental analysis with actual stereo images.
555 Comparison of Lithologic Mapping with ASTER,
Hyperion, and ETM Data in the Southeastern
Chocolate Mountains, USA
Xianfeng Zhang and Micha Pazner
An empirical comparison of the EO-1 Hyperion, EOS ASTER, and Landsat ETM sensors was performed to examine the utility of these three sensors for gold-associated lithologic mapping in the southeastern Chocolate Mountains area, California. Three images were evaluated with respect to three aspects: classification accuracy, matched filtering score index, and separability of the five significant rock types in the study area. The results show that the classifications from Hyperion and ASTER data are mostly similar with an overall accuracy of over 85 percent and kappa coefficient 0.81. Due to the presence of more SWIR and thermal bands, the Hyperion and ASTER images can achieve better lithologic mapping than ETM. The assessment of matched filtering score index and the separability also supports these findings. Hyperion can discriminate more similar classes than ASTER and ETM, while the better availability and spatial coverage makes the ASTER sensor more suitable for large-area lithologic mapping.
563 Use of Landsat ETM and Topographic Data to Characterize
Evergreen Understory Communities in Appalachian Deciduous Forests
Robert A. Chastain, Jr. and Philip A. Townsend
Evergreen understory vegetation was classified using Landsat ETM imagery and ancillary data in two physiographic provinces in the central Appalachian highlands; the Ridge and Valley and the Allegheny Plateau. These evergreen understory communities are dominated by rosebay rhododendron (Rhododendron maximum L.) and mountain laurel (Kalmia latifolia L.), which are spatially extensive and ecologically important to the structure and functioning of Appalachian forests. DEM-derived topographic information was integrated with Landsat data to assess its potential to improve classification accuracy, maximum likelihood, and minimum distance, and decision tree classification approaches were tested with these data in a factorial manner. An overall accuracy of 87.1 percent (Khat = .806) was achieved in the Ridge and Valley province by employing a maximum likelihood approach using Landsat data alone, while an 82.9 percent overall accuracy (Khat = .755) was obtained for the Allegheny Plateau employing a hybrid decision tree classification approach with Landsat and topographic data.
[figure 1.] [figure 3.] [figure 5.] [figure 6.]
577 An Object-oriented Approach to Urban Forest Mapping in Phoenix
Jason S. Walker and John M. Briggs
We present a classification procedure in order to delineate woody vegetation in an arid urban ecosystem using high-resolution, true-color aerial photography. We adopted an object-oriented approach due to the physical nature of high-resolution photography, in which the objects of interest were typically composed of many pixels. The segmentation process was parameterized to isolate vegetation patches from shrubs to large trees. These objects were then spectrally analyzed for discrimination between woody vegetation and all other objects and a classification scheme developed. Accuracy within subclasses was analyzed and indicated highest accuracy for large, dense vegetation. Error was caused by the following plant typologies: small vegetation, sparse canopy density, and gray vegetation. The outset of this procedure produces a binary matrix where the entire raster set is classified highlighting the elements of the urban forest.
585 Imagery Integration Methods for Precise Geological Mapping of Rugged Terrain, Alberta, Canada
Daniel Lebel, Guillaume Kenny, Donna Kirkwood, Jacynthe Pouliot, Jean-Sébastien Marcil, Christine Deblonde, and Patricia Molard
This paper reports on new geological mapping techniques using the rugged area of Moose Mountain, Alberta as a test site. First, we present a web-accessible photographic database that facilitates interactive visualization of multiple rock exposures for geology mapping, and second, the results of analysis of two readily available oblique photogrammetric methods, that have been previously applied to archeological and industrial surveys. These two techniques of high-resolution terrestrial oblique photogrammetry called “block bundle” and “terrain rendering” were tested using the web database, in order to evaluate their precision and accuracy as mapping techniques. The efficiency of the new techniques to support geological mapping is compared to other mapping techniques. The results show that in conditions where high-resolution digital elevation and imagery data are available, a web-based terrain rendering technique, combined with a web-accessible photographic database of rock exposures is advantageous to a wide range of geological analyses, including analog modeling of hydrocarbon reservoirs.