联合时空信息的全波形激光雷达目标特征探测方法及关键技术研究

Full-waveform LiDAR is a novel remote sensing technology, which has appeared with the ability to record and analyze the complex waveform of the backscattered signal echo. Thus, it can improve the capability of detection and identification of LiDAR in complex remote sensing environment, but it is also a challenge for full-waveform data processing and target characteristics extraction. Firstly, aiming to the complex characteristics of the detected targets can not be extracted sufficiently, based on the theory of radioactive transfer and Bidirectional Reflectance Distribution Function (BRDF), the relationship between multi-characteristics of targets and the space-time information (full-waveform and laser profile image) can be accomplished. It will provide a basis of target characteristics extraction and identification algorithms. Secondly, based on methods of priori knowledge, data assimilation and cooperative inversion, it is working out the problems about some disturbing factors of atmospheric turbulence, multi-scale transform and ill-posed inversion to modify the model. It can improve the robustness of the multi-parameters inversion model in a complex environment. Thirdly, for validating the methods mentioned above, a prototype of full-waveform LiDAR will be established with improvement. It will measure multi-level elevation and validate the multi-parameters inversion model. Lastly, we design the experiments, obtain the data, and modify the inversion model. We do performance analysis of full-waveform LiDAR systems, including the range accuracy, the false alarm probability, the detection probability and the noise threshold. This work can provide scientific basis for novel full-waveform LiDAR independently researched and developed by China.

全波形激光雷达是一种新概念遥感雷达手段，完整的激光回波信息可以提高复杂背景下目标特征的探测与识别能力，也对数据处理和目标特征探测方法提出了挑战。首先，本课题针对回波数据中目标特征不易解耦问题，结合辐射传输理论和双向反射率分布函数，综合利用全波形激光雷达的时空域信息（激光回波数据与激光轮廓信息），建立目标多参数反演模型，为研究目标特征提取、识别算法提供理论依据。其次，面向多种探测影响因素（大气湍流、尺度变换、病态反演），联合先验知识、数据同化及协同反演方法，修正目标多参数反演模型，提高复杂环境下反演模型的鲁棒性。再次，研制全波形激光雷达系统原理样机，完成多层次高程数据获取，验证目标多参数反演模型。最后，合理设计实验，完成数据采集，进行模型修正，实现目标多参数解算，分析系统性能（测距精度、虚警概率、探测概率和噪声阈值）。该研究成果可为全波形激光雷达系统的自主研发提供理论依据和验证手段。

激光雷达是基于卫星、飞机、汽车等观测平台，获取大范围、高精度、多层次空间信息的一种尖端综合技术，是解决国土资源观测、农林遥感信息获取、城市环境感知等一系列国计民生重大问题的现代技术手段。项目“联合时空信息的全波形激光雷达目标特征探测方法及关键技术研究”的主要研究内容是：①联合全波形激光雷达时空域数据，建立面向复杂目标特征探测的多参数反演模型，实现目标多参数解耦；②面向多种探测影响因素，进行多参数反演模型修正，提高系统设计的鲁棒性；③完成全波形激光雷达系统原理样机研制，实现了目标高精度三维成像与探测，为我国高精度载荷的自主研发提供验证手段和新思路。.本项目旨在研究全波形激光雷达及目标高精度三维成像。针对内容①计划，研究Richardson-Lucy反卷积分解法、经典高斯分解法和高斯渐进分解法；实验结果表明在低信噪比及高混叠情况下，高斯渐进分解法优于Richardson-Lucy反卷积分解法及经典高斯分解法；信噪比大于15dB时，点数提取粗差率小于10%，可实现厘米级的测距精密度。针对内容②计划，采用光线追踪法建立了五种安装角误差的测角误差模型，提出了网络法与双面法结合的模型参数估计方法；实现对系统安装角误差的自标定和点云校正。针对内容③计划，搭建了一套全波形激光雷达系统和实验验证平台；波形采样率为1.25GSa/s，探测范围5m-150m，测距精密度2cm，水平角分辨率0.8"，垂直角分辨率5.28'。研究快速、高精度全波形时刻鉴别方法，对推进我国的激光雷达探测成像技术具有重要作用和意义。.在国家自然科学基金面上项目资助下，2017-2020年发表论文21篇，其中SCI检索论文17篇，第一作者SCI检索论文6篇，EI检索期刊论文4篇；已经申请发明专利15项，已授权6项；参与发表论著2部。计划研究内容全部完成，达到预期目标。