机载激光雷达波形数据横向高斯分解方法

A Lateral Gaussian Decomposition Method for LiDAR Waveform Data

针对机载激光雷达波形数据分解易受噪声影响,高斯组分个数及叠加波初始参数估计不精确等问题,提出了一种横向高斯波形分解方法。该方法首先对波形进行滤波平滑处理,剔除背景噪声后,将检测到的波峰划分为不同的类型,分别估计其初始参数;然后横向逐步迭代分解估计初始高斯分量,在去除无效的初始高斯分量后,利用列文伯格-马夸尔特（Levenberg-Marquardt）算法进一步优化参数;最后解算得到分解点云。实验结果表明,该方法能有效地检测各种类型的回波信号,对叠加波形具有良好的适应性,并能在一定程度上保护弱波。相比系统点云,本文方法解算的点云在数量和细节上更具有优势,反映了更加丰富的地物垂直结构信息以及在森林参数获取方面的应用潜力。

The decomposition of waveform data is a key step in waveform analysis.Traditional waveform decomposition methods cannot detect overlapped sub-waveforms and weak sub-waveforms,and cannot appropriately estimate the number of Gaussian components.In this article,we propose a lateral Gaussian decomposition method.A waveform is smoothed after removing the background noise.We divide the detected waves into different types of waveforms,and estimate their initial parameters with different methods,then progressively laterally decompose waveform until all the Gaussian components are decided.After removing invalid components,we usethe Levenberg-Marquardt method to further optimize the parameters.Experiments show that this new method can effectively detect different kinds of complicated waveforms;demonstrating both robustness and efficiency.