基于IMU与激光雷达紧耦合的混合匹配算法

针对线、面特征匹配的激光雷达测距与地图构建算法(Lightweight and Ground-Optimized Lidar Odometry And Mapping,LeGO-LOAM)在自动导引运输车(Automated Guided Vehicle,AGV)室内室外实时建图与定位时,易出现激光里程计累积误差大和旋转估计不准确等问题,本工作采用惯性测量单元(Inertial Measurement Unit,IMU)与激光雷达紧耦合的LeGO-LOAM算法,通过IMU为激光雷达提供的初始位姿信息,构建IMU与激光雷达联合误差函数,实现位姿共同迭代优化.其中,对于室外结构化信息较少时,在点对点的迭代最近点算法(Iterative Closest Point,ICP)较高定位精度的基础上,结合LeGO-LOAM算法和ICP算法互补性,进一步提出基于IMU与激光雷达紧耦合的混合匹配算法:当环境中结构信息较多时,激光里程计采用LeGO-LOAM算法,而当环境中结构化信息较少时采用ICP算法.实验结果表明,基于IMU与激光雷达紧耦合的混合匹配算法可有效降低激光里程计相对位姿误差和累积误差,提高AGV小车定位精度以消除部分地图重影.     

Comparison of bathymetry and seagrass mapping with hyperspectral imagery and airborne bathymetric lidar in a shallow estuarine environment

We compared hyperspectral imagery and single-wavelength airborne bathymetric light detection and ranging (lidar) for shallow water (<2 m) bathymetry and seagrass mapping. Both the bathymetric results from hyperspectral imagery and airborne bathymetric lidar reveal that the presence of a strongly reflecting benthic layer under seagrass affects the elevation estimates towards the bottom depth instead of the top of seagrass canopy. Full waveform lidar was also investigated for bathymetry and similar performance to discrete lidar was observed. A provisional classification was performed with limited ground reference samples and four supervised classifiers were applied in the study to investigate the capability of airborne bathymetric lidar and hyperspectral imagery to identify seagrass genera. The overall classification accuracy is highly variable and strongly dependent on the classification strategy used. Features from bathymetric lidar alone are not sufficient for substrate classification, while hyperspectral imagery alone showed significant capability for substrate classification with over 95% overall accuracy. The fusion of hyperspectral imagery and bathymetric lidar only marginally improved the overall accuracy of seagrass classification.     

Assessment of the impacts of surface topography, off-nadir pointing and vegetation structure on vegetation lidar waveforms using an extended geometric optical and radiative transfer model

In order to prioritize the measurement requirements and accuracies of the two new lidar missions, a physical model is required for a fundamental understanding of the impact of surface topography, footprint size and off-nadir pointing on vegetation lidar waveforms and vegetation height retrieval. In this study, we extended a well developed Geometric Optical and Radiative Transfer (GORT) vegetation lidar model to take into account for the impacts of surface topography and off-nadir pointing on vegetation lidar waveforms and vegetation height retrieval and applied this extended model to assess the aforementioned impacts on vegetation lidar waveforms and height retrieval.Model simulation shows that surface topography and off-nadir pointing angle stretch waveforms and the stretching effect magnifies with footprint size, slope and off-nadir pointing angle. For an off-nadir pointing laser penetrating vegetation over a slope terrain, the waveform is either stretched or compressed based on the relative angle. The stretching effect also results in a disappearing ground peak return when slope or off-nadir pointing angle is larger than the “critical slope angle”, which is closely related to various vegetation structures and footprint size. Model simulation indicates that waveform shapes are affected by surface topography, off-nadir pointing angle and vegetation structure and it is difficult to remove topography effects from waveform extent based only on the shapes of waveform without knowing any surface topography information.Height error without correction of surface topography and off-nadir pointing angle is the smallest when the laser beams at the toward-slope direction and the largest from the opposite direction. Further simulation reveals within 20° of slope and off-nadir pointing angle, given the canopy height as roughly 25 m and the footprint size as 25 m, the error for vegetation height (RH100) ranges from − 2 m to greater than 12 m, and the error for the height at the medium energy return (RH50) from − 1 m to 4 m. The RH100 error caused by unknown surface topography and without correction of off-nadir pointing effect can be explained by an analytical formula as a function of vegetation height, surface topography, off-nadir pointing angle and footprint size as a first order approximation. RH50 is not much affected by topography, off-nadir pointing and footprint size. This forward model simulation can provide scientific guidance on prioritizing future lidar mission measurement requirements and accuracies.     

Physically based vertical vegetation structure retrieval from ICESat data: Validation using LVIS in White Mountain National Forest, New Hampshire, USA

Vegetation structure retrieval accuracies from spaceborne Geoscience Laser Altimeter System (GLAS) on the Ice, Cloud and land Elevation Satellite (ICESat) data are affected by surface topography, background noise and sensor saturation. This study uses a physical approach to remove surface topography effect from lidar returns to retrieve vegetation height from ICESat/GLAS data over slope terrains. Slope-corrected vegetation heights from ICESat/GLAS data were compared to airborne Laser Vegetation Imaging Sensor (LVIS) (20 m footprint size) and small-footprint lidar data collected in White Mountain National Forest, NH. Impact of slope on LVIS vegetation height estimates was assessed by comparing LVIS height before and after slope correction with small-footprint discrete-return lidar and field data.Slope-corrected GLAS vegetation heights match well with 98 percentile heights from small-footprint lidar (R² = 0.77, RMSE = 2.2 m) and top three LVIS mean (slope-corrected) heights (R² = 0.64, RMSE = 3.7 m). Impact of slope on LVIS heights is small, however, comparison of LVIS heights (without slope correction) with either small footprint lidar or field data indicates that our scheme improves the overall LVIS height accuracy by 0.4-0.7 m in this region. Vegetation height can be overestimated by 3 m over a 15° slope without slope correction. More importantly, both slope-corrected GLAS and LVIS height differences are independent of slope. Our results demonstrate the effectiveness of the physical approach to remove surface topography from large footprint lidar data to improve accuracy of maximum vegetation height estimates.GLAS waveforms were compared to aggregated LVIS waveforms in Bartlett Experimental Forest, NH, to evaluate the impact of background noise and sensor saturation on vegetation structure retrievals from ICESat/GLAS. We found that GLAS waveforms with sensor saturation and low background noise match well with aggregated LVIS waveforms, indicating these waveforms capture vertical vegetation structure well. However, waveforms with large noise often lead to mismatched waveforms with LVIS and underestimation of waveform extent and vegetation height. These results demonstrate the quality of ICESat/GLAS vegetation structure estimates.     

利用光子累计改善成像激光雷达的信噪比方法

介绍了采用光子计数探测的1064nm激光雷达系统。理论分析了光子计数探测技术和信噪比的改善，实验研究了改善信噪比的脉冲累计方法。实验给出了光子累计脉冲与信噪比之间的平方根关系，同时给出了不同距离下的探测结果，数据表明：当光子累计脉冲提高到原来的4倍时，信噪比达到平均1．89倍的增强。     

面向实际场景SLAM应用的光照适应性研究

为探究环境感知设备在SLAM算法应用过程中的光照适应性问题,在不同光照强度下分别进行激光雷达和深度相机SLAM算法的验证性评估实验.基于四轮差速机器人,搭载16线激光雷达和深度相机,结合LOAM(Lidar Odometry And Mapping)和RTAB-MAP(Real-Time Appearance-Based Mapping)算法,分别在明暗环境中分析验证设备光照适应性.实验结果表明:在明亮环境下,基于视觉SLAM和激光SLAM系统偏差的中误差分别为0.203和0.644 m;在黑暗环境中两者偏差的中误差分别为0.282和0.683 m;深度相机在明、暗环境中的定位建图效果均优于激光雷达,深度相机的光照适应性更强.     

三维激光雷达技术在输电线路运行与维护的应用

三维激光测量技术为空间三维信息的获取提供了全新的手段。介绍了三维激光雷达技术在输电线路运行与维护的应用,包括线路走廊危险地物检测、电力线间距离精细量测、输电线路三维可视化管理、线路走廊地形地貌变化检测、已有输电线路的增容分析和新建线路的树木砍伐评估,并给出激光雷达技术在我国输电线路运维中应用的实例。     

激光雷达测量大气气溶胶光学厚度方法研究

介绍一种激光雷达常数标定和气溶胶光学厚度(AOD)测量的新方法.利用太阳辐射计,获得大气气溶胶的光学厚度,激光雷达可以获得35～40 km高度的回波信号,在这一高度区间可忽略气溶胶的存在,大气模式可以提供大气分子散射系数,根据激光雷达方程计算出激光雷达常数.反之,标定激光雷达常数后,根据激光雷达方程,以激光雷达35～40 km的大气分子后向散射回波信号来确定气溶胶的光学厚度.激光雷达测量结果与太阳辐射计的测量结果一致性较好,说明该方法是可行的.这种新方法既可以用于白天的气溶胶光学厚度测量,也可以用于夜间测量.     

基于力传感的系留无人机定位方法研究

在无GPS或弱GPS环境下,系留无人机通常采用激光雷达或视觉进行定位,由于受距离和光线的影响,无法实现精准定位。针对该问题,提出了一种基于力传感的系留无人机绳缆定位方法。系留无人机在室内飞行时,绳缆的状态可等效于悬链线,通过建立系留无人机悬链线模型,可有效估计无人机所处空间位置。使用基于力传感器的无人机、硅胶绳缆、绕线机搭建测试平台,绳缆定位数据与MarveImind的Indoor“GPS”标定无人机的位置做误差对比分析。实验结果表明:基于力传感的系留无人机定位方法,相较于Fotokite Pro无人机绳缆定位方法,定位精度提升了70.86%。该定位方法的提出,有利于系留无人机在无GPS环境下的稳定飞行。     

应用独立分量分析的激光测风雷达湍流频谱分解

针对大型风力发电机机组中常见的脉动湍流、风机尾流与涡流等湍流信号,研究了利用自然梯度下降的独立分量分析方法的湍流频谱分离效果,以区分中心风速与湍流信号,提高风机机组的综合工作效率。首先分析了风机组中常见湍流信号的后向散射与频谱分布特点,然后依据这些特点设计了对应的独立分量分析模型。在仿真结果符合要求的基础上,进行了双目激光雷达天线的风速采集与实际分离效果检测。实验结果表明,在大气折射率结构常数C2n≤10-14同时广义大气常数α≥4的通常情况下,利用双目信号能够分离出一个湍流中心和一个中心风速。对1 s内两个谱峰的波动范围进行统计,获得(2.59±0.05)MHz的中心风速以及(1.22±0.19)MHz的湍流中心估计,且二者的平均信噪比分别为25.93 dB和31.01 dB,能够在获得稳定的中心风速估计的同时得到一个较为稳定的湍流中心估计。