We proposed a hybrid imaging scheme to estimate a high-resolution absolute depth map from low photon counts. It leverages measurements of photon arrival times from a single-photon LiDAR and an intensity image from a c...We proposed a hybrid imaging scheme to estimate a high-resolution absolute depth map from low photon counts. It leverages measurements of photon arrival times from a single-photon LiDAR and an intensity image from a conventional high-resolution camera. Using a tailored fusion algorithm, we jointly processed the raw measurements from both sensors and output a high-resolution absolute depth map. We scaled up the resolution by a factor of 10, achieving 1300 × 2611 pixels and extending ~4.7 times the unambiguous range. These results demonstrated the superior capability of long-range high-resolution 3D imaging without range ambiguity.展开更多
山体滑坡会导致生命和财产损失,获取完整的滑坡空间分布图及对易发区域的准确判定有利于指导生产、生活和生态空间优化。在滑坡调查过程中,茂密的植被覆盖使滑坡调查难度加大,机载激光雷达(light detection and ranging,LiDAR)技术的穿...山体滑坡会导致生命和财产损失,获取完整的滑坡空间分布图及对易发区域的准确判定有利于指导生产、生活和生态空间优化。在滑坡调查过程中,茂密的植被覆盖使滑坡调查难度加大,机载激光雷达(light detection and ranging,LiDAR)技术的穿透能力使真实地形特征得以呈现,从而实现植被茂密区滑坡识别。该文通过仿地飞行获取研究区LiDAR点云数据,基于点云数据得到数字高程模型(digital elevation model,DEM),在山体阴影分析、彩色增强显示及三维场景模拟基础上,识别出区域内已有滑坡的位置与规模,经野外核实,滑坡解译精度为86.4%。针对滑坡易发区评价问题,以现有滑坡为样本,首次采用遥感分类思维开展滑坡易发区划定,采用小区域内与滑坡发育有关的高程、坡度和地表起伏度组合成影像,以支持向量机为分类方法,判定出滑坡易发区域,经滑坡检验样本分析,滑坡识别精度为81.91%。研究表明:基于高精度的LiDAR数据及其视觉增强后的图像能识别小型滑坡,采用支持向量机分类法可以准确确定滑坡易发区,为下一步三生空间规划与优化提供依据。展开更多
针对现有基于伪点云的3D目标检测算法精度远低于基于真实激光雷达(Light Detection and ranging,LiDar)点云的3D目标检测,本文研究伪点云重构,并提出适合伪点云的3D目标检测网络.考虑到由图像深度转换得到的伪点云稠密且随深度增大逐渐...针对现有基于伪点云的3D目标检测算法精度远低于基于真实激光雷达(Light Detection and ranging,LiDar)点云的3D目标检测,本文研究伪点云重构,并提出适合伪点云的3D目标检测网络.考虑到由图像深度转换得到的伪点云稠密且随深度增大逐渐稀疏,本文提出深度相关伪点云稀疏化方法,在减少后续计算量的同时保留中远距离更多的有效伪点云,实现伪点云重构.本文提出LiDar点云指导下特征分布趋同与语义关联的3D目标检测网络,在网络训练时引入LiDar点云分支来指导伪点云目标特征的生成,使生成的伪点云特征分布趋同于LiDar点云特征分布,从而降低数据源不一致造成的检测性能损失;针对RPN(Region Proposal Network)网络获取的3D候选框内的伪点云间语义关联不足的问题,设计注意力感知模块,在伪点云特征表示中通过注意力机制嵌入点间的语义关联关系,提升3D目标检测精度.在KITTI 3D目标检测数据集上的实验结果表明:现有的3D目标检测网络采用重构后的伪点云,检测精度提升了2.61%;提出的特征分布趋同与语义关联的3D目标检测网络,将基于伪点云的3D目标检测精度再提升0.57%,相比其他优秀的3D目标检测方法在检测精度上也有提升.展开更多
基金supported by the Key-Area Research and Development Program of Guangdong Province (No.2020B0303020001)the National Natural Science Foundation of China (Nos.62031024 and 12104443)+5 种基金the Innovation Program for Quantum Science and Technology (No.2021ZD0300300)the Shanghai MunicipalScienceandTechnologyMajorProject (No.2019SHZDZX01)the Shanghai Science and Technology Development Foundation (No.22JC1402900)the Shanghai Academic/Technology Research Leader (No.21XD1403800)the Shanghai Sailing Program (No.21YF1452600)the Natural Science Foundation of Shanghai (No.21ZR1470000)。
文摘We proposed a hybrid imaging scheme to estimate a high-resolution absolute depth map from low photon counts. It leverages measurements of photon arrival times from a single-photon LiDAR and an intensity image from a conventional high-resolution camera. Using a tailored fusion algorithm, we jointly processed the raw measurements from both sensors and output a high-resolution absolute depth map. We scaled up the resolution by a factor of 10, achieving 1300 × 2611 pixels and extending ~4.7 times the unambiguous range. These results demonstrated the superior capability of long-range high-resolution 3D imaging without range ambiguity.