POINT CLOUD GENERATION FROM GAUSSIAN DECOMPOSITION OF THE WAVEFORM LASER SIGNAL WITH GENETIC ALGORITHMS

Abstract

Abstract: Recent developments in LIDAR technology lead to the availability of the waveform systems, which capture and digitize the whole return of the emitted LASER pulse. As many objects may cause multiple returns in the same echo, one task is to detect and separate different echoes within the same digitized measurement. In this paper the results of a study aimed at LASER signal waveform decomposition using genetic algorithms are introduced. The proposed method is based on the Gaussian decomposition approach and analyzes each digitized return to compute one or more points. Initially, the number of peaks contained in the waveform is determined by a simple peak detection method, with a local maximum point algorithm. When more than one peak is detected, genetic algorithms are applied to estimate the amplitude, time and standard deviation of each peak within the digitized signal. With this methodology it was possible to increase the number of points by approximately 17 % compared to the point cloud obtained using commercial software. The best results were obtained in areas with high vegetation, and thus the methodology can be applied to the generation of denser points cloud in forest areas.