Influence Of Terrain Gradient On Mean Canopy Height Retrieve Of Airborne LiDAR

Lidar is an active remote sensing technology,which can accurately obtain three-dimensional information of surface objects,With the development of lidar system and storage technology,more and more airborne small spot lidar systems have the ability to acquire waveform data,that is,to acquire independent point cloud data and waveform data at the same time.Waveform data is unprocessed original echo sequence,and point cloud data is generated directly by hardware system.Therefore,the application of lidar in forestry is further promoted.The slope caused by the complex fluctuation of the earth's surface have a certain broadening effect on the waveform.The change of terrain will result in two situations affecting the forest echo waveform of LiDAR:?Topographic fluctuations change the distance between forest components and sensors,and even cause the overlapping effect between the ground echoes on the uphill and the forest echoes on the downhill;? Slopes increase the actual area of the speckle and contain more trees than flat forests under the same conditions.In this study,Inner Mongolian Irogen district was taken as the research area.By reading airborne full-waveform LiDAR data and pre-processing such as initial parameter estimation of denoising and smoothing,waveform fitting and so on,all waveform data in the sample were superimposed and processed to simulate pseudo-large spot waveform data.DEM was obtained by filtering airborne LiDAR point cloud data,and terrain factors of slope was extracted from DEM by combining GIS software.Finally,combined with the prior knowledge of large spot data processing,the influence of forest topography on pseudo-large spot waveform data was analyzed,which mainly includes two aspects:one was to simulate the influence of different slope topography on echo signal;the other was to verify the effect by using pseudo-large spot waveform data.The main research contents and results were as follows:(1)Waveform fitting of pseudo-large spotDue to the spot diameter of small footprint LiDAR is too small,the canopy height can not be represented by the pulse first echo.To lower the impact and further improve the application potential of small footprint airborne LiDAR full-waveform data in forest structural parameter estimation.In this study,a new method based on Gaussian decomposition by waveform fitting was proposed.Additionally,the airborne LiDAR full-waveform data of Inner Mongolian Irogen district was obtained and processed.By calculating the ratio of the Gaussian component pulse energy to total pulse energy in the sample,it was regarded as the corresponding weights of each Gaussian component feature parameter,and the weighted average of the amplitude,position and half-wave width of the feature parameter was obtained respectively,which was the the feature value corresponding to the Gauss function of pseudo large footprint waveform data.Based on the waveform data and the pseudo large footprint data extraction characteristic parameters,the regression model was established by combining the field measured Lorey 's average tree height respectively,and the comparative analysis was carried out.The model determination correlation R2 of the waveform inversion model was 0.47,the average overall precision P was 78.19%,the model determination correlation R2 of the pseudo large footprint inversion model was 0.61,and the average overall precision P was 90.65%.It was concluded that the inversion accuracy of the pseudo large footprint model is higher than the waveform inversion model,which reduces the effect brought from the small footprint diameter of LiDAR,and excavates the application potential of the airborne LiDAR waveform data of small spot.(2)The influence of slope on waveformFiltering is the key step of airborne LiDAR point cloud data processing.This research was based on the random forest method to filter.Four criteria,i.e.class I error,class II error,total error and kappa coefficient,were used to evaluate and analyze.The results showed that the random forest method can achieve high-precision classification of point cloud data,which is of great significance to improve the automation and intelligence level of airborne LiDAR point cloud classification.Digital Terrain Model(DEM)was acquired based on point cloud filtering results,and slope information was extracted from DEM with GIS software.Because of the complex undulation of the earth's surface,the terrain in the laser radar spot is not flat and uniform in many cases,often contains certain undulation.There are great differences in slope in different regions.In this study,the three-dimensional lidar forest echo waveform model was used to simulate the effect of terrain on waveform,and the simulation results were verified by the pseudo-large spot waveform data of lidar.The results showed that the slope has a great influence on the waveform of pseudo-large spot lidar.With the increase of the slope,both the ground echo and the canopy echo broaden,and the waveform length increases.At the same time,the wave peaks of the ground and vegetation decreased,and the echoes from the ground decreased significantly,and gradually overlapped with the echoes near the ground.(3)Influence of slope on mean canopy height retrieve of pseudo large spotIn order to establish the estimation model of mean canopy height,this study extracted characteristic parameters based on pre-processed small spot waveform data and pseudo large spot waveform data,drawed scatter plots of waveform parameters and measured average tree height,observed their correlation,then determined the regression model of average tree height,and established model evaluation index for model construction and accuracy evaluation.Then,the correlation analysis of the mean canopy height and the extracted slope results based on pseudo large spot inversion was carried out,and the effect of slope on the mean canopy height was further analyzed.It was found that the mean canopy height increased with the increase of slope.