全波形机载激光雷达绝对辐射定标与不确定性分析

为揭示全波形激光雷达回波在森林等植被区域多回波信号的特征和对目标识别分类的影响，以激光雷达方程为模型基础，利用朗伯体目标为地面参考，提出了将激光雷达波形参数标定为后向散射截面、后向散射系数和漫反射率等物理量的方法，实现了机载小光斑全波形机载激光雷达数据绝对辐射定标。对两个不同实验区的LMS-Q680i数据标定结果表明，漫反射率与参考反射率相对误差总体分别小于10%和5.5%，误差标准差分别为0.044和0.077，有效消除了条带间差异。推导了多回波的激光雷达方程组，比较了相同系统在不同观测条件下的定标常数变化，重点分析了全波形激光雷达在穿透性目标上的多回波现象造成的能量衰减，及其对辐射定标结果的影响，证明了多回波现象是造成多回波信号减弱的主要原因。该现象是当前技术体制下激光雷达观测过程本身存在的缺陷，对基于激光雷达辐射信息的目标识别分类带来了一定的挑战，也是多光谱、高光谱激光雷达辐射信号定标必须解决的问题。

Uncertainty analysis of the absolute radiometric calibration of full waveform airborne LiDAR

The multiple return phenomenon of laser pulse in the technology of full waveform laser scanning has an inevitable impact on the radiometric signal of LiDAR data as well as their radiometric calibration. Previous studies have been made that characterized this phenomenon as an attenuation of laser pulse energy partially intersected by objects, such as canopy and building edges, through the travel path. In this study, we proposed a novel theoretical explanation of multiple returns by establishing a set of LiDAR equations, one for each sub-pulse that composed the original outgoing pulse. The energy attenuation of LiDAR signals through penetrable targets, such as forest canopy, and its influence on radiometric calibration were particularly analyzed. Comparative experiments were conducted with data from one laser instrument of Riegl LMS-Q680i LiDAR system in two different data collection campaigns. One data collection site was covered by LiDAR flight lines of 600 m and 1200 m Above Ground Level (AGL), and the other site with all 600 m AGL. During the data acquisition, ground earth surface with approximate Lambertian reflectivity behavior were measured with filed spectrometer, and the reflectance of ground reference objects were applied in the radiometric calibration process. The data were processed and radiometrically calibrated on the basis of classical LiDAR equation. In addition, multiple return point cloud of a scene with homogeneous ground surface with planted vegetation were extracted for further quantitative analysis. This process was implemented to reveal and characterize the influence of multiple returns on full waveform LiDAR echoes and subsequent target classification. Through the quantitative comparison of data strips, deviations of overlapping data of different flying altitude were calculated. It was demonstrated from the results that the systematic data deviations of LiDAR strip parameters are successfully eliminated. The overall relative errors of corrected diffuse reflectance of the two regions are less than 10% and 5.5%. The standard deviations of strip difference are 0.044 and 0.077 accordingly. Calibration constants in independent LiDAR surveying campaigns are compared. The constants were found to be with correlation to the LiDAR system and flying configurations. Moreover, it was found that LiDAR returns of different return number were not consistent, despite that they were reflected by the same object surface. Significant weakening was observed in the returns of higher orders. It was concluded that multiple return is the major cause of return intensity weakening on homogeneous surfaces and it has crucial effects on radiometric information based target recognition. This problem cannot be readily solved with the current LiDAR observation mechanism in typical mapping scenarios. Challenges from this phenomenon are inevitable to further target recognition and should be addressed for advanced multiple and hyperspectral LiDAR data in the future.