Waveform Processing of Laser Pulses for Reconstruction of Surfaces in Urban Areas

In this paper we describe investigations for digital recording of received laser pulses and a detailed analysis of the pulse waveform. In contrast to previous works using a single photon detection technique, we now focus on multi photon detection. An experimental system was constructed using an airborne laser scanning system and a receiver unit wih an appropriate sampling rate for a fast recording of laser pulses. Capturing the temporal waveform by simultaneous scanning of the scene delivers a data volume (spacetime cube). Such a data cube of the received signals allows exploiting different features without selecting a special feature or a measuring technique in advance. For the main investigation a ground based sensor platform is used to measure the urban scene. Surface variations within the spatial beam width result in special waveforms. We explored the capabilities of analyzing the shape of the waveform for surface features in form of distance, roughness, and reflectance for each backscattered pulse. An iterative estimation algorithm (Gauss-Newton method) is proposed to get a parametric description of the original waveform by a Gaussian. Single responses and more than one response (multiple pulses) are investigated. Additionally, the data cube opens the possibility of considering neighborhood relations for space-time filtering or segmentation. Image slices extracted from the data cube were processed to estimate line segments for surface features. This has a significant advantage: the estimated lines are not based on single range points received from a pulse detector which delivers more or less accurate range values, but they are based directly on the intensity distribution of the backscattered signal. Since many values of the intensity distribution have contributed to each line segment, the accuracy of location can be much better than the pixel dimension of the image slice.