ABSTRACT The objective of segmentation on point clouds is to spatially group points with similar properties into homogeneous regions. Segmentation is a fundamental issue in processing point clouds data acquired by LiDAR and the quality of segmentation largely determines the success of information retrieval. Unlike the image or TIN model, the point clouds do not explicitly represent topology information. As a result, most existing segmentation methods for image and TIN have encountered two difficulties. First, converting data from irregular 3-D point clouds to other models usually leads to information loss; this is particularly a serious drawback for range image based algorithms. Second, the high computation cost of converting a large volume of point data is a considerable problem for any large scale LiDAR application. In this paper, we investigate the strategy to develop LiDAR segmentation methods directly based on point clouds data model. We first discuss several potential local similarity measures based on discrete computation geometry and machine learning. A prototype algorithm supported by fast nearest neighborhood search and based on advanced similarity measures is proposed and implemented to segment point clouds directly. Our experiments show that the proposed method is efficient and robust comparing with algorithms based on image and TIN. The paper will review popular segmentation methods in related disciplines and present the segmentation results of diverse buildings with different levels of difficulty.

The ISPRS Foundation is an independently registered philanthropic entity that has been established SOLELY for providing financial assistance to advance the benevolent purposes of the International Society for Photogrammetry and Remote Sensing (ISPRS), an international, non-profit, non

Extraction of individual tree crown objects in urban areas using available VHR space-borne imaging systems remains a challenging image processing task mainly because of two factors: the limited spectral information offered by the sensors and their limited spatial resolution. The spectral information of available VHR satellite sensors is not sufficient to discriminate tree crowns from other land cover classes such as grassland and shrubs using spectral pixel-based classification techniques. This can be solved by using a contextual classification approach, while the limitation of spatial resolution can be reduced by using Super resolution mapping (SRM). In this paper we extend the contextual Markov Random Field (MRF) based SRM method developed earlier for multispectral images to include the panchromatic band with higher spatial resolution. We apply this method for extraction of tree crown objects from Quickbird images, setting the spatial resolution of superresolved map to that of the panchromatic band, 0.6 m. We apply the proposed method for identification of tree crown objects in a residential area in the Netherlands. We find that incorporation of panchromatic band leads to improved object identification. Our object based accuracy assessment indicate that the proposed method identified 73 % of the trees in the study area, with some commission errors in tree areas with understory vegetation. 1

ABSTRACT As one of the advanced analysis capabilities in GIS, network analysis provides strong decision support for users in searching optimal path, finding the nearest facility and determining the service area. To lead to an effective solution for GIS network analysis, a new random searching method --genetic algorithm is introduced and applied in this article. The classical combinatorial optimization problem (knapsack problem) is used to introduce the concepts of genetic algorithm and describe its various operations in detail, such as encoding, crossover, mutation and inversion. Selection of parameters to reach the optimal performance for the genetic algorithm is also discussed in general. The GIS network analysis is then formalized as a general combinatorial optimization problem consisting of an objective function and a general constraint condition. The implementation details of this algorithm are discussed in terms of encoding and genetic operations. Test results for a network with eighty nodes are presented to demonstrate the efficiency of the genetic algorithm and its application potential.

GIS and digital mapping operations frequently require the automatic comparison and superimposition of geometric figures represented by sets of vertex coordinates supported by structural and topological information. When the configurations are not structured, that is the only vertex coordinates of the figures are available, manual intervention is needed in order to establish correspondences among the different geometries. To overcome this limitation, an automatic method has been developed to detect the correspondences between two or more equivalent sets of unlabeled points, representing n-dimensional geometric figures. The proposed technique performs a geometrical analysis of the adjacency matrices of the point configurations, in order to identify, for each one, the vertex of maximal asymmetry. A pairwise comparison of the sorted components of the adjacency matrix relative to these vertices, leads to the identification of the point correspondences. A directly-computed Procrustes conformal transformation is then applied to the geometric figures in order to achieve their optimal alignment. Also in case of geometric entities included into another, the problem solution starts trying to find some minimal asymmetric subconfigurations (kernels) that are similar in both figures. A Procrustes superimposition of these corresponding kernels is then applied, and extended to the remaining points of the included configuration. A shape test is finally executed in order to identify the best solution. Specific geometric rules and filters are implemented to optimise the computation process. The method has been successfully tested on cadastral cartographic matching problems. In addition, it is suitable for a wider range of possible applications, like CAD/CAM, computer vision and reverse engineering. 1.

Commission V – WG 6 In this paper we address all the problems and solutions of converting a measured point cloud into a realistic 3D polygonal model that can satisfy high modeling and visualization demands. Close range photogrammetry deals since many years with manual or automatic image

Now we are developing a wearable recording system which can capture images from user’s view point in user’s everyday life automatically. The user refers to the images which is acquired using this system later, and these images support a user’s activity, such as human memory and human thoughts. Amount images which are acquired becomes so huge that it is difficult for the user to refer them. So the mechanism for viewing effectively is very important, such as summary. In this research, we describe the concept for summary mechanism of everyday life images. 1. INTRODUNCTION We aim at developing the system, which can record people's everyday life as information and reuse the information, to be able to support people's memory and thoughts[1][2][3][4]. The behavior image of the everyday life is acquired and recorded as a view-point image. This system has a mechanism which it can acquire images selectively (Figure 1). 2. EVERYDAY LIFE RECORDING AND GIVING ATTRIBUTED INFORMATION 2.1 Everyday life recording system

Photogrammetric data, whether it refers to terrain or to close-range images, are typically massive in size. This results in a dataset that is often too large to handle and prohibits any form of real-time interaction or even visualization. Therefore, special database structures and image storing formats should be sought. FlashPix image format supports images of any size, and any resolution. Images in FPX format are stored at multiple independent resolutions. This special structure allows applications to select the appropriate resolution for a certain procedure and to access directly the specific areas of an image needed for the operation being performed. This eliminates the need to process the entire image for viewing a small part of it, or to handle a large amount of data (high resolution image) in order to produce a smaller amount of data (displaying a lower resolution image). This paper presents research efforts on identifying visualization tools, which can make use of this special image format, taking thus advantage of its special data structure. 1.

Commission V- WG 6 In the valley of Bamiyan, Afghanistan, ca 2000 years ago, two big standing Buddha statues were carved out of the sedimentary rock of the region. They were 53 and 35 meters high and the Great one figured as the tallest representations of a standing Buddha. In March 2001 the Taleban militia demolished the colossal statues, as they were considered an insult to Islam. After the destruction, a consortium was established to rebuild the Great Buddha of Bamiyan at original shape, size and place. Our group did the computer reconstruction of the statue, which can serve as basis for the physical reconstruction. In this paper we report the results of the 3D reconstruction of the Great Buddha, with particular attention to the modeling and visualization of the measurements.

The goal of an augmented reality system (ARS) is to superimpose in real time a real world scenery with a virtual extended version of itself. Such an ARS is, also developed as part of a disaster management tool of the collaborative research centre 461 (CRC461): "Strong Earthquakes". Rescue units are supposed to use the ARS as a tool to plan their actions on site using the possibilities offered by virtual reality. Regarding the reconnaissance strategy of the CRC461 there will be airborne laser scanning data collected of the whole affected area after an earthquake. This in turn means that the geometrical shape of the buildings is known. This threedimensional data can be fused with other information available, e.g., digital elevation model, building structure and so on. This information can now be used as planning information for rescue units. The construction, of such a system, is a challenge in many ways. Firstly, the proposed work shows how digital surface models can be used in different ways for the on site calibration of an See-Through Head-Mounted Display (STHMD) and the connected head tracking devices. Next to this the paper will introduce special possibilities offered by such a technology to analyse possible rescue plans for collapsed buildings in the context of disaster relief. On the other hand there is a set of primitive analysis techniques like measuring distances without touching the object directly. 1.