Author_Institution :
Sch. of Instrum. Sci. & Opto-Electron. Eng., Beihang Univ., Beijing, China
Abstract :
In this paper, a novel decomposition method of full-waveform Light Detection And Ranging (LiDAR) echo is proposed. This decomposition method includes four steps. In the first step, the noise mean value and deviation are estimated, meanwhile noise threshold and fitting threshold are configured. In the second step, full-waveform echoes are smoothed using the Gaussian kernel function. In third step, maximum points and inflection points are detected aiming at filtered echoes in the second step. Meanwhile, the number of echo component in full-waveform echo and initial parameters of each echo component are estimated based on the maximum points and inflection points, and echo components are sorted depending on the area of echo components. In the fourth step, the full-waveform echoes are one by one addition fitted using Levenburg-Marquardt (LM) method based on sorted result and fitting threshold. At last, the correct rate, and relative and absolute error of this decomposition method are simulation verified using MATLAB software based on randomly generated data. The result shows that the whole decomposition correct rate of this method is 82%, the maximal relative error of location is 0.18%, the maximal absolute error of amplitude is 1.2157 millivolts, the maximal absolute error of location is 0.5548 nanoseconds, and the maximal absolute error of FWHM is -2.219 nanoseconds. The decomposition correct rate and parameter error show that decomposition coefficients can be further used for target feature extraction and classification.
Keywords :
Gaussian processes; echo suppression; feature extraction; filtering theory; geophysical signal processing; optical radar; remote sensing by radar; signal classification; smoothing methods; Gaussian kernel function; LM method; Levenburg-Marquardt method; Matlab software; decomposition coefficients; decomposition correct rate; echo component sorting; filtered echoes; fitting threshold; full-waveform LiDAR echo decomposition method; full-waveform echo smoothing; inflection points; initial echo component parameter estimation; light detection-and-ranging echo; maximal absolute FWHM error; maximal absolute amplitude error; maximal absolute location error; maximal relative location error; maximum points; noise deviation estimation; noise mean value estimation; noise threshold; parameter error; randomly generated data; simulation verification; target feature classification; target feature extraction; Accuracy; Fitting; Laser radar; Measurement by laser beam; Noise; Remote sensing; Surface emitting lasers; Gaussian smoothing; full-waveform LiDAR; nonlinear least squares; waveform fitting;