Study On LiDAR Waveform Simulation Model And Waveform Quantifications

Light Detection And Ranging(LiDAR)is one of the state-of-art remote sensing(RS)techniques for Earth observation.With the advantage of direct,quick,and accurate acquisition of 3D spatial information,LiDAR has shown great potentials in many applications,e.g.,forest investigation,topography surveying,electric power inspection,and digital city.To make better use of the LiDAR technique,it is necessary to find out the physical relationship between surface radiational and structural characteristics and LiDAR observations and to further improve the quantification of LiDAR RS applications.LiDAR signal simulation models can simulate the laser-surface interaction mechanisms that give rise to the LiDAR signals.It is a celebrated tool to solve many crucial problems of LiDAR quantitative remote sensing,such as sensor design,quantitative inversion of surface parameters,etc.As far as the current LiDAR systems are concerned,full-waveform LiDAR measures the complete profile of a return signal by sampling it in fixed time intervals,with theoretically unlimited number of measurements.The acquired waveform comprehensively reveals surface vertical distribution within the footprints.Discrete return and photon-counting LiDARs can be regarded as products of further sampling of the full-waveform data.Based on the above background,this study conducted the LiDAR waveform simulation model and waveform quantification researches.The main objectives are i)to establish LiDAR waveform simulation model as a basic tool for follow-up works;ii)to conduct a series of sensitivity analyses of effects of multiple factors on LiDAR forest waveform;iii)to propose the methods of the design and verification of the LiDAR sensor parameters;iv)to propose quantitative inversion models of surface structural characteristics.The main works and conclusions of this study are as follows:(1)This study proposed a new LiDAR waveform simulation model named DART-Lux.Previous LiDAR waveform simulation models don't explain clearly the LiDAR remote sensing mechanism in a physical manner and show low efficiency when modeling LiDAR signals over complex large-area scenes.DART-Lux is an efficient LiDAR waveform simulation model based on the physical-based light transfer principles.Specifically,it establishes the 3D Earth landscapes with facets,turbid medium,or their combination;a laser source model and a LiDAR receiver model are simulated to imitate a real LiDAR sensor;the advantaged computer-graphics algorithms(e.g.,bidirectional path tracing,multiple importance sampling,direct light sampling,geometry instance)is reimplemented in LiDAR radiative transfer process and redesigned to simulate the LiDAR power and distance measurements.By comparing with the existing LiDAR model,DART-Lux shows higher simulation accuracy(R~2=1,r RMSE=0.21%),higher efficiency(time-consuming reduced by more than half),and smaller memory usage(hundredfold less memory).The comparison with actual LiDAR waveforms also indicated DART-Lux can achieve accurate simulation of LiDAR waveform signal(R~2=0.88,RMSE=0.0016)for various land covers.(2)This study established a LiDAR forest scattering component model and explored the scattering mechanisms of the forest canopy.Specifically,we integrated the LiDAR waveform simulation model DART-Lux,the leaf reflectance model PROSPECT,and the forest growth model TASS to simulate the LiDAR forest waveforms.Then a LiDAR forest scattering component model was established to explore the laser scattering mechanism in the forest canopy by quantifying the ratio of the multiple scattering signal to the total waveform.The results indicated that the laser multiple scattering in the canopy enhances the waveform intensity;the forest scattering components received in different time delays depend on the vertical distribution of surface objects;the ratio of the multiple scattering signal to the total received waveform is influenced by laser wavelength,footprint diameter,forest structural characteristics,leaf biochemical characteristics,topography geometry,and so on.(3)This study proposed a method of design of large-footprint full-waveform LiDAR footprint size and a method of evaluation of footprint horizontal geolocation accuracy of spaceborne full-waveform LiDAR.The method of designing footprint diameters quantifies the requirements of forest and topography applications and then imported them into the LiDAR waveform simulation model as the limiting conditions.The results concluded that the footprint diameter for forest applications should be 10.6m?25.0 m,and for topography applications should be less than 32.3 m.The method of evaluating footprint horizontal geolocation accuracy is proposed based on the waveform simulation model.It matches the real LiDAR waveform with the DSM data.The best matching position is considered the real footprint center and is used to calculate the footprint horizontal offset.Results show that this method is only applicable for the areas with significant height characteristics,e.g.,urban.In the urban area,the method indicates the horizontal geolocation accuracy of the GLAS footprints is about 8.19 m.(4)This study proposed a series of quantitative inversion models of surface structural characteristics.For surface terrain,a terrain slope estimation model is proposed using spaceborne full-waveform LiDAR data.This model considers the influence of the footprint shape,orientation,size,and terrain aspect on the terrain slope estimation.Compared with the slope estimation methods without considering these factors,the estimation accuracy of terrain slope is improved by about 15%.For the forest canopy,a forest leaf area index(LAI)inversion model is proposed by quantitatively combining full-waveform LiDAR data and medium-resolution optical images.This model corrects the between-canopy clumping effect that causes the underestimation of forest LAI.The inversion accuracy is RMSE=0.39,R~2=0.83.For the city buildings,a method of extraction of multiple building heights is proposed using full-waveform LiDAR data assisted by high-resolution optical images.This method achieves the accurate extraction of multi-target three-dimensional information at the sub-footprint scale.The extraction accuracy is R~2=0.97,r RMSE=13.2%by verification with field measurements.