EXTRACTION OF FEATURES FROM OBJECTS IN URBAN AREAS USING SPACE- TIME ANALYSIS OF RECORDED LASER PULSES B. Jutzi a , U. Stilla b a FGAN-FOM Research Institute for Optronics and Pattern Recognition, 76275 Ettlingen, Germany - jutzi@fom.fgan.de b Photogrammetry and Remote Sensing, Technische Universitaet Muenchen, 80290 Muenchen, Germany - stilla@bv.tum.de Commission II, WG II/2 KEY WORDS: Urban, Analysis, Simulation, Laser scanning, LIDAR, Measurement, Feature. ABSTRACT: In this paper we describe a simulation system and investigations for analysis of recorded laser pulses. A simulation setup that considers simulated signals of synthetic objects was developed for exploring the capabilities of recognizing urban objects using a laser system. The temporal waveform of each laser pulse is analyzed for gaining the pulse properties: range, pulse power and number of peaks. Considering the received pulse power of the associated spatial neighborhood for the region boundary delivers the estimation of the edge position and edge orientation with sub pixel accuracy. 1. INTRODUCTION The automatic generation of 3-d models for a description of man-made objects, like buildings, is of great interest in photogrammetric research. In photogrammetry a spatial surface is classically measured by triangulation of corresponding image points from two or more pictures of the surface. The points are manually chosen or automatically detected by analyzing image structures. Besides this indirect measurement using object characteristics, which depends on natural illumination, active laser scanner systems allow a direct and illumination- independent measurement of the range. Laser scanners capture the range of 3-d objects in a fast, contactless and accurate way. Overviews for laser scanning systems are given in (Huising & Pereira, 1998; Wehr & Lohr, 1999; Baltsavias, 1999). Current pulsed laser scanner systems for topographic mapping are based on time-of-flight ranging techniques to determine the range of the illuminated object. The time-of-flight is measured by the elapsed time between the emitted and backscattered laser pulses. The signal analysis to determine the elapsed time typically operates with analogous threshold detection (e.g. peak detection, leading edge detection, constant fraction detection). Some systems capture multiple reflections, caused by objects which are smaller than the footprint located in different ranges. Such systems usually capture the first and the last backscattered laser pulse (Baltsavias, 1999). Currently first pulse as well as last pulse exploitation is used for different applications like urban planning or forestry surveying. While first pulse registration is the optimum choice to measure the hull of partially penetrable objects (e.g. canopy of trees), last pulse registration should be chosen to measure non-penetrable surfaces (e.g. ground surface). Figure 1a shows a section of an image taken in first pulse mode. The foliage of the trees is visible. Figure 1b was taken in last pulse mode. The branches and foliage are not visible and the building areas are smaller than in Figure 1a. Due to multiple pulse reflection at the boundary of the buildings and the processing by first or last pulse mode, building areas dilate or erode. For visualizing the various sizes of the building footprints in first and last pulse images a difference image is calculated (Figure 1c). The ambiguous pixels of the building are visible by a bright area along the buildings contours. A zoomed section of a building corner is depicted in Figure 1d. Building edges are expected within these bright areas. a b c d Figure 1. Sections of an urban scene (Test area Karlsruhe, Germany). a) elevation images captured by first pulse mode, b) elevation images captured by last pulse mode, c) difference image of first and last pulse mode, d) subsection of the difference image (building boundary).