基于无人机遥感影像拓扑分析的三维重建

在无需任何地面控制点或其它先验知识前提下,探索一种基于无人机遥感影像的三维重建方法.利用无人机飞控数据建立的影像拓扑结构,依次通过特征点提取、影像匹配、从运动恢复结构等步骤估计出相机位置和姿态参数,并恢复出场景特征点云信息,最后对重建精度进行分析.试验结果表明,文中提出的方法可快速、可靠地实现较高精度的三维模型重建.     

2D地图的建筑物场景结构提取方法及其在城市增强现实中的应用

城市AR中,场景结构的自动获取,对自适应信息表达具有重要意义,是解决“层叠式”信息表现形式引发的信息指示不明、场景感知混乱等问题的关键步骤。然而,场景结构信息隐藏在场景图像中,直接提取十分困难。2D地图描述了地理实体的位置、轮廓和空间布局,可作为场景结构提取的先验信息。针对城市AR中场景结构获取难、效率低的问题,本文提出一种基于2D地图的建筑物场景结构自动化提取方法。在地理配准的基础上,首先根据2D地图中建筑物的轮廓和属性信息,构建场景图像语义分割的结构线索;然后将其作为过分割处理后场景图像合并的依据,生成多个包含语义信息的图像区域;最后根据图像区域与2D地图中建筑物轮廓之间的映射关系,提取出场景图像中的区域轮廓、场景深度和平面朝向等场景结构信息。试验选取了格拉茨地区32个建筑物场景进行测试。结果表明本文方法能实时地提取建筑物场景结构,且建筑物立面提取的质量明显优于对比方法。提取的场景结构能应用于城市AR多种自适应信息表达方法中,使信息标注更明确,遮挡关系更清晰。     

高分六号宽幅相机在轨几何定标及精度验证

高分六号宽幅相机能够实现单相机成像幅宽优于800 km,对大尺度地表观测和环境监测具有独特优势。在轨几何定标是光学遥感卫星几何处理的关键环节,直接影响影像的几何质量。本文充分考虑高分六号宽幅相机超大视场的畸变特性以及多谱段的成像特点,提出宽幅相机在轨几何定标方法,采用基于探元指向角的几何定标模型补偿宽幅相机系统误差,通过绝对定标和相对定标方法联合估计各波段的内外定标参数。利用Landsat 8影像、资源三号DSM为参考数据,对宽幅相机进行绝对定标处理,再利用ASTER GDEM为参考数据进行相对定标处理,其几何定标结果表明,高分六号宽幅影像绝对定位精度在3像素左右,内部几何精度能稳定在1像素,且波段间配准精度在0.3像素以内,表明在轨几何定标后高分六号宽幅影像几何质量得到了明显的提升。     

稀少控制的多平台星载SAR联合几何定标方法

几何定标采用地面控制点获取距离-多普勒模型中的精确几何参数,用于完成星载SAR影像高精度几何定位。但在广域范围内,特别是高山地区域,控制点极难获取。此外,传统定标方法仅面向单一平台SAR影像,尚不能实现多平台影像的联合几何定标。针对上述问题,本文提出一种基于稀少控制的多平台星载SAR联合几何定标方法。该方法从包含实测控制点的主影像出发,使用点位追踪算法获取主影像与从影像之间的连接点,并以连接点为桥梁逐级完成从影像的几何定标。本文采用京津冀地区南北向分布共计235 km的3景TerraSAR-X、3景TanDEM-X、5景高分三号影像进行联合几何定标试验,仅使用5个控制点即完成了所有影像的几何定标,并利用SF-3050星站差分GNSS接收机采集实测GPS点进行精度评价。结果表明使用稀少控制点定标后的TSX/TDX影像的几何定位精度优于3 m,GF-3影像的几何定位精度优于7.5 m,验证了该方法的有效性和正确性。     

A hybrid measurement approach for close-range photogrammetry

Full automation in close-range photogrammetric measurement has long been a practical reality. However, constraints apply to the process such that automated 3D measurement is generally confined to targeted points in an environment of controlled illumination. The ready availability of consumer-grade digital cameras has made photogrammetric measurement accessible and more widely employed for a host of new applications, the majority of which call for 3D measurement of other than signalized object features. Process automation is therefore typically precluded. This paper discusses a hybrid measurement approach which involves fully automatic network orientation with targets, while at the same time supporting follow-up semi-automatic and manual operations such as feature point and line extraction and surface measurement via image matching. The topics discussed include camera calibration, the metric exploitation of colour attributes, issues related to image point correspondence determination, operator assisted feature measurement and surface extraction. All are important to the practical realisation of the hybrid measurement approach.     

面阵图像定位精度提升方法研究

为了提升面阵相机成像定位精度,以吉林省松原市北部区域的机载面阵图像为实验对象,对成像过程进行误差分析与标定. 机载面阵相机成像存在由三个姿态角引起的定位误差以及由相机镜头畸变等因素产生的畸变误差. 针对以上问题,首先采用严密几何模型,得到面阵图像初步成像结果;然后,采用后方交会和多项式附加参数模型等方法解决内、外方位元素引起的误差问题,得到较为精确的定位结果;最后进行实验验证,测试结果表明,经过内、外方位元素的误差定标,能够控制在5 m以内的范围,精度提升94.37％,处理效果显著. 通过研究标定面阵成像过程中内、外方位元素产生的误差,提高了面阵相机定位精度,对面阵相机成像的应用和推广具有一定价值和科学参考依据.      

Using horizontal and vertical building structure to constrain indirect sensor orientation

This paper presents a method to integrate linear horizontal, vertical and right-angled scene structures into the bundle adjustment of image sequences. An increasing number of airborne image acquisition systems is available and equipped with non-metric small- or medium-frame cameras and no or insufficiently accurate INS devices. In cases where the data is to be used for the production of geo-spatial data, where a certain accuracy and precision is required, an indirect sensor orientation, possibly including self-calibration, needs to be performed. The idea which led to the presented approach is to reduce the number of GCPs necessary for this task by applying the mentioned scene structures. The method directly uses the linear structures, visible at man-made objects as fictive observations within the adjustment, while self-calibration of intrinsic camera parameters and lens distortion is included as well.Experiments with two datasets demonstrate that, through this method, only limited GCP information is required to obtain satisfactory results. In fact, in one experiment using oblique images, several scene constraints were provided and only the datum was defined by ground control. The residuals at check points from this setup were similar to the traditional case where several well-distributed GCPs were available in the scene. In the second experiment the ability of this approach to support the bundle adjustment was shown for a UAV dataset. Although no GCP and camera calibration information was available, the visual inspection of adjusted object points and the residuals at horizontal structures confirmed the ability of the method to align an image block with the structure, as embodied in the defined scene constraints. Despite the convincing outcome of the experiments, it needs to be mentioned that some manual work is still involved in defining the constraints. In future work the issue of automation will be addressed.     

光学遥感影像道路提取的方法综述

道路信息在多个应用领域中发挥着基础性的作用。光学遥感影像能够以较高的空间分辨率对目标地物进行精细化解译,可大幅增强地物目标的提取能力。充分利用光学遥感影像丰富的几何纹理信息,进行道路的精确提取,已成为当前遥感学界研究的热点与前沿问题。鉴于此,本文依据近年来大量相关文献,对现有的理论与方法进行了归类与总结,通过分析不同方法采用的道路特征组合,将道路提取方法划分为模板匹配、知识驱动、面向对象和深度学习4类方法,简要介绍了道路提取普适性的评价指标并对部分方法进行了分析与评价;最后对现有光学遥感影像道路提取的发展提出了建议和展望。     

图像分层匹配的HJ-1A/B CCD影像自动几何精校正技术与系统实现

环境与灾害监测预报小卫星星座(环境一号卫星,HJ-1A/B)自发射以来,在环境监测、灾害评估、土地资源调查等领域发挥了重要的作用。但是HJ-1A/B卫星CCD图像的2级产品(HJ-1 CCD图像)几何精度低,实际应用中需要进行几何精校正。HJ-1 CCD图像具有宽覆盖、大视场角、几何变形复杂的特点,几何精校正难度大。针对该问题,本文提出了一个以Landsat TM全球拼接图像为基准,基于Forstner算子和模板匹配的分层配准方法。该方法使用分层匹配获得的大量高精度且分布均匀的控制点构建Delaunay三角网,有效地解决了HJ-1 CCD图像的几何精校正问题。在配准技术研究的基础上,研发了HJ-1 CCD图像几何精校正系统,系统具有全球HJ-1 CCD图像的自动批量处理能力。实验结果表明,本文提出的几何精校正方法精度高,实现了环境星图像的自动批量处理。     

Image sequence based automatic multi-camera system calibration techniques

An image sequence-based, fully automatic and rather flexible procedure for the calibration of stationary multi-camera systems for 3-D observation of dynamic events is presented and analysed. While conventional close-range camera calibration techniques are either based on a stable point-field with known reference coordinates or on a temporarily stationary point-field with only approximately known 3-D coordinates, which is imaged from different locations and under different orientations with one single camera, the presented technique is based on stationary cameras and moving targets, making use of the image sequence acquisition nature of most solid state cameras. In the simplest version, only one easily detectable marker has to be tracked through image sequences of multiple pre-calibrated cameras, thus avoiding the necessity of homologous feature identification for the establishment of multi-view correspondences; 3-D coordinates of the marker position are not required. This single-marker method does not allow for the determination of the interior orientation. In an extended version allowing for full camera orientation and calibration, a reference bar of known length is moved through object space, with the problem of feature identification and establishment of multi-view correspondences being reduced to the tracking of two targets. The method can only be used with multi-camera systems and is most useful for 3-D motion analysis applications, but may be adapted to a wide range of other applications. The advantages of the method over conventional self-calibration techniques are the trivial establishment of multi-view correspondences, the fact that no temporarily stable target field has to be constructed, and the fact that each camera has to be set up only once. After an explanation of the technique, its performance is examined in detail based on extensive computer simulations, and the practical effectiveness is shown in a pilot study on industrial robot calibration. Based on these studies, recommendations are given concerning the number of reference bar locations, preferable reference bar orientation schemes and the achievable accuracy potential.     

Innovations in Automation for Vision Metrology Systems

The advantages of vision metrology (VM) systems over their film-based counterparts, in terms of the degree of automation within the measurement process, were initially confined to the image mensuration stage. Of late, however, the potential of significantly raising the level of automation of offline and real time VM systems has been more fully exploited, to the point where full automation of the offline procedure has been realized. Innovative features such as automatic image point identification via coded targets, exterior orientation through automatically recognizable "control" groupings of target points, determination of image point correspondences within unlabelled point clouds, and real time tracking of multipoint, handheld measurement probes are now incorporated in industrial VM systems. Another significant innovation is the development of "smart" cameras with incorporated image processing and measurement capabilities which remove the necessity for transmitting image data to the host computer. This paper reviews recent innovations in automation and discusses their operational impact on practical VM applications. The concepts discussed are illustrated through reference to a modern VM system which supports offline and real time measurement.     

Automatic detection of a one dimensional ranging pole for robust external camera calibration in mobile mapping

A mobile mapping system (MMS) is the answer of the geoinformation community to the exponentially growing demand for various geospatial data with increasingly higher accuracies, captured by multiple sensors. As the mobile mapping technology is pushed to explore its use for various applications on water, rail, or road, the need emerges to have an external sensor calibration procedure that is portable, fast and easy to perform. This way, sensors can be mounted and demounted depending on the application requirements without the need for time consuming calibration procedures. A new methodology is presented to provide a high quality external calibration of cameras which is automatic, robust and fool proof. The method uses a portable, standard ranging pole which needs to be positioned on a known ground control point. While the literature focuses on solving the absolute orientation problem of the calibration, an automatic method to detect the calibration object is missing. Here, we present a mutual information based image registration technique for automatic sub-pixel localization of the ranging pole under realistic outdoor conditions. We include several robust error detection rules to allow the calibration to run without human intervention, giving as little outliers as possible, to ensure a correct calibration. Several tests are performed under various lighting and noise conditions which prove the methodology’s robustness.     

Evaluation of the geometric stability and the accuracy potential of digital cameras — Comparing mechanical stabilisation versus parameterisation

Recent tests on the geometric stability of several digital cameras that were not designed for photogrammetric applications have shown that the accomplished accuracies in object space are either limited or that the accuracy potential is not exploited to the fullest extent. A total of 72 calibrations were calculated with four different software products for eleven digital camera models with different hardware setups, some with mechanical fixation of one or more parts. The calibration procedure was chosen in accord to a German guideline for evaluation of optical 3D measuring systems [VDI/VDE, VDI/VDE 2634 Part 1, 2002. Optical 3D Measuring Systems–Imaging Systems with Point-by-point Probing. Beuth Verlag, Berlin]. All images were taken with ringflashes which was considered a standard method for close-range photogrammetry. In cases where the flash was mounted to the lens, the force exerted on the lens tube and the camera mount greatly reduced the accomplished accuracy. Mounting the ringflash to the camera instead resulted in a large improvement of accuracy in object space. For standard calibration best accuracies in object space were accomplished with a Canon EOS 5D and a 35 mm Canon lens where the focusing tube was fixed with epoxy (47  μm maximum absolute length measurement error in object space). The fixation of the Canon lens was fairly easy and inexpensive resulting in a sevenfold increase in accuracy compared with the same lens type without modification. A similar accuracy was accomplished with a Nikon D3 when mounting the ringflash to the camera instead of the lens (52  μm maximum absolute length measurement error in object space). Parameterisation of geometric instabilities by introduction of an image variant interior orientation in the calibration process improved results for most cameras. In this case, a modified Alpa 12 WA yielded the best results (29  μm maximum absolute length measurement error in object space). Extending the parameter model with FiBun software to model not only an image variant interior orientation, but also deformations in the sensor domain of the cameras, showed significant improvements only for a small group of cameras. The Nikon D3 camera yielded the best overall accuracy (25  μm maximum absolute length measurement error in object space) with this calibration procedure indicating at the same time the presence of image invariant error in the sensor domain. Overall, calibration results showed that digital cameras can be applied for an accurate photogrammetric survey and that only a little effort was sufficient to greatly improve the accuracy potential of digital cameras.     

Design and implementation of a computational processing system for off-line digital close-range photogrammetry

Through the adoption of recent innovations in automation in vision metrology, it can be demonstrated that rigorous, yet user-friendly digital photogrammetric processes of calibration and orientation/triangulation can be incorporated into a computational scheme which is on the one hand capable of meeting the demands of high metric quality, while on the other offering the facilities necessary to support wider application by non-specialist users. The software system Australis, developed for image mensuration and restitution of off-line close-range photogrammetric networks, is featured to illustrate these processes and procedures. By describing the structure and components of Australis, the authors aim to demonstrate that many processes which have on occasion been viewed to be the exclusive province of automated, high-precision vision metrology are indeed suited to more general application across a broad range of fields which involve 3D object recording via close-range imagery.     

特征提取策略对高分辨率遥感图像场景分类性能影响的评估

高分辨率遥感图像场景分类方法主要涉及两个环节:特征提取以及特征分类,分类器的设计已经相对成熟,当前工作的重点是特征提取策略的研究。为了进一步推动特征提取策略的研究,将特征提取策略对高分辨率遥感图像场景分类性能的影响进行了定性和定量评估。首先,回顾了高分辨率遥感图像场景分类的发展历程;然后,对现有高分辨率遥感图像场景分类方法的特征提取策略进行分类总结,并从理论上将各类特征提取策略对场景分类性能的影响进行定性评估;最后,在3个规模较大的数据集上对多种特征提取策略进行实验对比,将不同特征提取策略对场景分类性能的影响和各数据集的复杂度进行定量评估。     

控制点识别的全自动相机标定算法研究

针对现有相机标定环节中控制点需要手动选取,标定自动化程度不高,效率较低的问题。该文在圆心投影点精确定位研究成果的基础上,提出了一种基于角距的控制点自动识别算法,解决了三维高精度控制场下全自动相机标定的关键环节。初步实验表明,在同等条件下,基于该方法的全自动标定精度、效率和稳定性均优于传统人工选取控制点的标定方法。     

基于图像语义分割的动态场景下的单目SLAM算法

在恢复场景信息和相机运动时,传统的SLAM算法是基于静态环境假设的。场景中的动态物体会降低算法的稳健性和最终的定位精度。本文提出将基于深度学习的图像语义分割技术与传统的视觉SLAM算法结合,以减少动态物体对定位结果的干扰。首先,构建有监督的卷积神经网络对输入图像中的动态物体进行分割,获得语义图像;然后,从原始图像中提取特征点,并根据语义图像剔除动态物体特征点,保留静态物体特征点;最后,利用静态物体特征点采用基于特征点的单目视觉SLAM算法对相机运动进行跟踪。在ApolloScape自动驾驶数据集上的试验表明,与传统方法相比,本文算法在动态场景中定位精度提升约17%。     

高分三号SAR影像广域范围联合几何检校技术

在获取广域范围雷达遥感正射影像图时,需要对区域内每景SAR影像进行几何参数修正,提高SAR影像的定位精度。但在广域范围进行控制点的选取和影像的逐一修正,其工作量巨大,严重影响了SAR影像分析应用的效率。对此,本文提出了一种广域范围联合几何检校方法。该方法在R-D模型的基础上,首先利用少许SAR影像计算其独立的系统级检校参数,其次去除系统级检校参数中大气延迟分量的影响,然后依据最小二乘平差原理获取最优的系统级检校参数,最后利用最优的系统级检校参数对所有SAR影像进行几何检校处理。论文采用覆盖中国中东部地区的32景GF-3SAR影像进行联合几何检校试验,检校后的SAR影像定位中误差优于9 m。试验结果表明:该方法能够提高GF-3 SAR影像几何检校的精度,验证了其有效性和可行性。     

Orthorectification of VHR optical satellite data exploiting the geometric accuracy of TerraSAR-X data

Orthorectification of satellite data is one of the most important pre-processing steps for application oriented evaluations and for image data input into Geographic Information Systems. Although high- and very high-resolution optical data can be rectified without ground control points (GCPs) using an underlying digital elevation model (DEM) to positional root mean square errors (RMSEs) between 3 m and several hundred meters (depending on the satellite), there is still need for ground control with higher precision to reach lower RMSE values for the orthoimages. The very high geometric accuracy of geocoded data of the TerraSAR-X satellite has been shown in several investigations. This is due to the fact that the SAR antenna measures distances which are mainly dependent on the terrain height and the position of the satellite. The latter can be measured with high precision, whereas the satellite attitude need not be known exactly. If the used DEM is of high accuracy, the resulting geocoded SAR data are very precise in their geolocation. This precision can be exploited to improve the orientation knowledge and thereby the geometric accuracy of the rectified optical satellite data. The challenge is to match two kinds of image data, which exhibit very different geometric and radiometric properties. Simple correlation techniques do not work and the goal is to develop a robust method which works even for urban areas, including radar shadows, layover and foreshortening effects. First the optical data have to be rectified with the available interior and exterior orientation data or using rational polynomial coefficients (RPCs). From this approximation, the technique used is the measurement of small identical areas in the optical and radar images by automatic image matching, using a newly developed adapted mutual information procedure followed by an estimation of correction terms for the exterior orientation or the RPC coefficients. The matching areas are selected randomly from a regular grid covering the whole imagery. By adjustment calculations, parameters from falsely matched areas can be eliminated and optimal improvement parameters are found. The original optical data are orthorectified again using the delivered metadata together with these corrections and the available DEM. As proof of method the orthorectified data from IKONOS and ALOS-PRISM sensors are compared with conventional ground control information from high-precision orthoimage maps of the German Cartographic Survey. The results show that this method is robust, even for urban areas. Although the resulting RMSE values are in the order of 2-6 m, the advantage is that this result can be reached even for optical sensors which do not exhibit low RMSE values without using manual GCP measurements.