多组分特征融合的小光斑全波形激光雷达隐蔽目标提取方法研究

Extraction of concealed targets in the vegetation-covered area has an important significance in the field of emergency rescue, environmental monitoring, urban safety management and so on. Compared to bare targets, only little information from the concealed targets can be obtained because of the shading from surrounding environment, which leads to the serious information interlaced aliasing phenomenon between adjacent targets, and greatly increases the difficulty of targets detection and information extraction. It is difficult to extract the concealed targets information with high-precision based on the traditional LiDAR. Small footprint full-waveform LiDAR can record the whole returned waveform while obtaining the information of three-dimensional coordinates of ground points, which has unique advantages especially to concealed targets extraction, and the properties of targets can be obtained by processing waveform data. Therefore, our project intends to take advantage of full-waveform LiDAR to develop a new way to extract the concealed targets. Firstly, we are going to build a laser radiative transfer model under the condition of vegetation coverage. Secondly, a waveform decomposition algorithm for the detection of the missing weak or compositive component will be proposed. Then the component features that are in accord with typical concealed targets will be explored. At last, an original concealed targets extraction method will be developed, which combines spatial information and physical attributes. And field experiment will be conducted to verify the accuracy of concealed targets information extraction. This project is the continuation and expansion of the existing research results. The related research is a sophisticated exploration in the field of active LiDAR remote sensing; undoubtedly it will make a huge contribution to the applications of earth observation by LiDAR technology in our country.

植被覆盖条件下的隐蔽目标提取在应急减灾、环境监测、城市安全管理等领域具有非常重要的意义。相比于裸露目标，隐蔽目标由于周围环境对目标的遮蔽仅可获取较弱信息量，进而导致相邻目标间信息交错混叠效应更加严重，大大增加了目标探测及信息提取的难度。现有基于传统激光雷达的提取方法难以进行隐蔽目标的高精度信息提取，小光斑全波形激光雷达在获取地面点三维坐标信息的同时可记录整个回波波形，对隐蔽目标提取有其独特优势。因此，本项目拟利用回波波形经处理可获得目标丰富属性特征的特点，构建有覆盖条件下的激光辐射传输模型，提出顾及弱波和叠加波遗失组分探测的波形分解算法，挖掘符合典型隐蔽目标的组分特征，发展一种融合空间信息和物理属性的多特征隐蔽目标提取方法，结合野外试验验证目标信息提取精度。本项目是申请人已有研究成果的延续和拓展，相关研究是激光主动遥感技术领域的深层探索，对提升我国激光雷达对地观测的应用潜能具有重要价值。

针对植被覆盖条件下的弱信息量隐蔽目标难以提取的瓶颈问题，项目充分利用全波形激光雷达能穿透植被获取隐蔽目标波形信息的独特优势，发展了一种融合空间信息和物理属性的多特征隐蔽目标提取方法。项目开展了植被覆盖条件下的激光雷达辐射传输机理研究，构建了最优形状参数支持的回波波形建模、提出了发射脉冲辅助的自适应噪声阈值波形去噪方法、自迭代判断波形拟合的顾及遗失组分的波形分解方法、联合改进局部G系数和回波率的多特征融合隐蔽目标提取方法。项目基于野外实测机载全波形激光雷达数据进行了典型隐蔽目标提取试验，验证了多特征融合隐蔽目标提取方法的有效可行性。.由于隐蔽目标返回的回波数量相比于裸露目标少，且大多是弱波和叠加波信号，为保证后向散射回波波形中的隐蔽目标组分能够被完整的探测和分解提取，项目从回波波形去噪、波形分解到目标提取的多个环节完成了算法探索。.项目创新性提出一种自适应噪声阈值估计联合AICC准则的全波形激光雷达波形分解方法，采用联合发射脉冲波形数据、顾及相邻采样值强度关联性的滤波去噪思路，消除背景噪声与随机噪声；对存在的小样本、弱回波隐蔽目标数据，优选AICC准则辅助完成波形分解，弥补了基于传统阈值法滤波去噪并结合AIC准则存在分解适应性差的不足；在传统有限混合模型（FMM）的列文伯格-马夸尔特（LM）波形分解算法的基础上，增加联合最小惩罚距离匹配（PMMD），保障弱波和叠加波信号的完整提取。此外，项目新增获得了多种波形组分特征，创新改进了空间统计分析的自相关局部G系数，实现了疑似隐蔽目标区域的高效探测，最终保障了典型隐蔽目标的有效提取。.本项目研究成果初步展现了全波形激光雷达数据在隐蔽目标提取方面的应用潜力，有助于服务于植被覆盖条件下汽车等典型隐蔽目标的提取，为隐蔽目标提取提供了一种新的解决途径。本项目研究发展的方法，亦可应用于其他地物目标的探测和提取，在应急减灾、城市安全管理等方面都具有很好的应用前景。