| The photon counting laser altimetry adopts single photon detectors as receiving devices.Benefiting from the high sensitivity and repetition rate of single photon detectors,photon counting laser altimetry could detect the continuous elevation of vegetation canopy and earth surface.Although the spaceborne photon counting laser altimetry plays an important role in forest structure mapping and improving global digital terrain model for its extremely high elevation accuracy and along orbit resolution,the single photon events that randomly occurr according to the detection probability make the point cloud data uncertain.Especially for vegetation detection,the single photon detection system not only brings the ability of continuous canopy elevation detection and vertical structure detection,but also brings new difficulties to its data processing,including: The density and distribution of signal point cloud which is vulnerable to the interference of background noise are variable,that increases the difficulty of filtering;The density of vegetation point cloud is relatively low,as well as the detection probability in monopulse detection,and then,the time of the first single photon event in monopulse detection is not necessarily the starting position of pulse echo,but more inclined to a period of time with higher detection probability,which may cause an error in canopy height inversion;In order to have enough information to retrieve the parameters of vegetation,we need to increase the size of sample plot window,which may cause errors due to the change of ground slope and uneven distribution of vegetation in the window.At present,there are few data processing algorithms for spaceborne photon counting laser altimetry vegetation point cloud.More optimized point cloud filtering algorithms,more accurate parameter inversion algorithms are needed to successfully process the spaceborne photon counting laser altimetry vegetation point cloud and extract the corresponding forest parameters from the point cloud.Based on the above background,the following work is carried out in this paper:(1)Through the statistics of ATLAS(the Advanced Topological Laser Altimetry System)point cloud data,we analyze the characteristics of vegetation point clouds with different times,seasons and species of vegetation.Then,a horizontal directional parameter is proposed to describe the continuity of the point cloud,and provides a certain theoretical basis for the parameter selection of subsequent data processing algorithms.(2)Based on the characteristics of spaceborne photon counting laser altimetry vegetation point cloud,an extremums searching model for rough filtering,and an adaptive directional density clustering model for fine filtering are propsed.By searching the position of canopy and ground in the height histogram of the point cloud,the rough filtering model effectively determines the elevation range of single points,and obtains the prior information to guide the subsequent filtering.The fine filtering algorithm adaptively changes the density threshold,neighborhood radius and continuous direction in the density clustering algorithm according to the characteristics of the object point cloud,which can effectively filter out the noise points and retain the signal points belonging to the tree crown and ground.These models are expected to solve the filtering difficulty caused by the noisy vegetation point clouds,the density of point cloud with large variation,the changing continuity in differ environment parameter situations.(3)Based on the characteristics of single photon point cloud,a high-precision canopy height retrieval method is proposed.Combined with the density and elevation percentage of point cloud,an improved TIN(Triangulated Infrared Network)model is utilized to classify the points of canopy top and ground.And then,a specific contour was obtained by interpolation to inverse canopy height.The accuracy of canopy height inversion is verified by comparing the vegetation canopy heights obtained by LIDAR(Light Detection and Ranging)data.At last,a quantitative error estimation model that could estimate and correct the average canopy height error caused by single photon detection mechanism is put forward based on the optical transmission model of vegetation and the single photon detection theory.(4)The leaf area index(LAI)is extracted from a new LAI inversion algorithm that take advantage of photon counting laser altimetry system’s ability of detecting the vegetation’s vertical structure.Based on the single photon detection probability,the relationship between the sample plot window size and the signal number deviation is calculated,and the basis for selecting the window size in inversion is analyzed.Then,the number of vegetation points and ground points are corrected to meet the requirement of energy ratio for the relation of points number and echo energy are not linear.;According to the ground contour,the influence of the slope which can not be ignored under large window size is corrected;The inversion error of LAI caused by the uneven distribution of vegetation in sample plot window is analyzed based on canopy optical transmission model of vegetation.The change of canopy extinction capacity caused by the uneven vegetation distribution is difficult to be corrected directly by adjusting the energy representend by signal points.Therefore,the new algorithm divides the signal point cloud once again,and obtains LAI by fitting the attenuation amplitude of energy representend by signal points in each layer and the attenuation amplitude of theoretical energy.The accuracy of this algorithm is verified by simulated point cloud for the airborne LIDAR data also has the error caused by the uneven distribution of vegetation.Faced with the new problems of data processing and and parameters inversion of spaceborne photon counting laser altimetry,this paper proposes more effective methods of filtering and vegetation parameters inversion.The object-oriented directional adaptive filtering method could accurately filter out the noise of vegetation point cloud in different environments,and provide reliable signal point cloud for subsequent vegetation parameter inversion;The canopy height inversion algorithm and its error estimation method can successfully extract the crown and ground contour and calculate the inherent system error caused by single photon detection system;The two LAI inversion methods which could accurately retrieve LAI,solve the problems in LAI inversion with spaceborne photon counting point cloud,including the nonlinearity of point numbers and echo energy,the surface slope and uneven vegetation distribution.These data processing and parameter inversion methods have solved the key technical difficulties in the data processing and vegetation parameters inversion with spaceborne photon counting laser altimetry point cloud,which lays a solid foundation to monitor the growth of forest vegetation by spaceborne photon counting laser altimetry. |
PDF Full Text Request