基于深度学习的异源立体影像对匹配方法

相对于同源影像立体匹配,基于无人机倾斜摄影与近景摄影获取的异源影像在空间特征、视场角及分辨率等方面均存在较大的差异,给影像匹配带来困难。本文利用基于单应性变换的卷积神经网络提取特征点,在匹配阶段采用交叉注意力机制的图神经网络进行特征点匹配。该方法较好地克服了异源影像间因存在较大视差和扭曲变换而导致的匹配效果较差的问题,并以河北省廊坊市大城县的马家祠堂为试验数据,对比传统SURF(加速稳定性征)算法与深度学习算法的匹配效果。结果表明,基于深度学习算法对存在大视角差异的异源影像的匹配率更高。     

基于交叉点的道路曲线化简算法研究

现有的曲线化简算法不能很好地化简具有交叉路口的道路曲线,针对这一问题提出一种基于交叉点的道路曲线化简算法。算法分为预化简和修正化简两个阶段:首先识别并得到曲线上的分段点,利用相邻的分段点作为道格拉斯-普克算法的首尾点对曲线进行化简,得到预化简的结果;然后对于交叉点引入偏差阈值ε,通过判断道路曲线交叉点与化简后交叉点的距离与偏差精度ε的大小关系来确定该交叉点的化简与保留,如果保留或者化简后的道路曲线没有交叉点那么将原交叉点作为分段点对此段曲线进行重新化简。理论分析与实验结果表明,文中算法能够有针对性地保留或化简道路交叉点以及保持曲线化简后的形态特征。     

计算机视觉中基于多照片的同名点自动匹配

对多张图像上同名点的自动匹配算法进行了分析,提出了利用多张图像上的核线约束来实现同名点的匹配。首先,计算出一个待匹配点在另外两张图像中的核线,在核线附近确定出概略匹配点。然后,再次利用核线约束特性在确定的概略匹配点之间确定出唯一的匹配点。最后,用该方法对多张图像间的同名点进行了匹配试验,实验结果表明,该方法效果良好,具有较好的实用价值。     

计算机视觉中基于多照片的同名点自动匹配

对多张图像上同名点的自动匹配算法进行了分析,提出了利用多张图像上的核线约束来实现同名点的匹配.首先,计算出一个待匹配点在另外两张图像中的核线,在核线附近确定出概略匹配点.然后,再次利用核线约束特性在确定的概略匹配点之间确定出唯一的匹配点.最后,用该方法对多张图像间的同名点进行了匹配试验,实验结果表明,该方法效果良好,具有较好的实用价值.     

一种基于嫦娥一号CCD影像的影像匹配方法

月球测绘是完成月球探测任务的基础保障。这里提出了一种适合于CE-1获取的CCD影像的多尺度约束自动匹配方法。首先利用SURF算子提取影像上特征点;然后进行基于准核线和最小欧式距离约束的影像匹配;最后采用随机采样算法对误匹配点进行剔除而得到同名点信息。实验结果表明,该匹配方法提取的同名点有利于CE-1月球影像DEM的生成。     

ERRORS IN RESECTION

Abstract

I. Although surveying is generally conducted on the principle that there should be more observations than are mathematically necessary for the calculation of the required results, the determination of the position of a point by resection from three fixed points is an exception to the principle. In observing a round of angles at a point, one of them can be inferred, but it is always observed in order to obtain a check by closing the round; a point fixed by intersections is generally observed from four or five surrounding points, although only two shots are strictly necessary. But the problem of resection from three points is a “minimum-of-data” problem, that is, the observations are only just sufficient to fix the point, and there is nothing extra to give a check on the result.     

基于岭估计的有理多项式参数求解方法

在使用最小二乘法解算卫星遥感影像的RPC参数时,如果控制点非均匀分布或模型过度参数化,其法方程系数矩阵很容易产生病态,获得的解将偏离真值,甚至得到错误的解.使用岭估计可改善法方程的状态,保证解稳定.采用岭估计方法,通过所获取的不同岭参数对SPOT和QuickBird影像进行实验,证实L曲线法是一种稳定的、有效的岭参数确定方法,可显著提高RPC参数的解算精度.     

一种基于语义分簇聚类的P2P空间数据索引机制

分布式环境下空间数据的索引是空间数据处理中一个关键性的基础问题,引入了控制点及四叉树划分结构,并通过Hash函数把控制点映射到Chord网络中,在此基础上提出了基于语义的分簇聚类分布式四叉树的空间数据索引机制(spatial data index based on clustering distributed quad-tree,SDI-CDQT),该机制主要包含四叉树划分、空间数据查询和分簇聚类3个子算法。实验表明,SDI-CDQT机制是可行和有效的。     

区域GPS形变监测网的数据处理方法比较分析

获得合理的形变分析量是形变监测的前提。本文以测站坐标和速率为参数,分析比较了形变分析数据处理的3种方法:静态平差解、动态平差解以及滤波解,由于随机误差的存在,这3种解变得不等价;进而分析了这3种数据处理的方差分量解。最后以陕西省中部的一形变监测网为例,分析了其应用效果。     

基于最小二乘配置的九参数模型在三维坐标转换中的应用

运用传统的布尔莎七参数法进行三维坐标转换时,并未考虑到两坐标系统的公共点坐标误差及三个坐标轴缩放比例不均的问题。引入最小二乘配置法,将公共点误差作为随机信号处理,并增加两个尺度参数解决坐标轴缩放比例不均的问题;推导出基于最小二乘配置的九参数模型及其误差方程,通过实验对比分析表明,该方法可以有效地提高坐标转换精度。     

一种适用大旋转角的GPS网三维平差方法

在工程测量中往往采用工程坐标系,该坐标系坐标轴可能与GPS采用的WGS84坐标系的相应坐标轴间存在较大的旋转角,此时若仍直接按照一般方法进行三维约束平差计算会给结果带来不利影响,本文给出了一种适用大旋转角的GPS网三维平差方法,该方法能够很好地处理大旋转角时直接采用一般三维平差模型计算带来的不利影响,同时又能够获得平差后工程网中各点的点位精度。工程应用表明该方法非常有效,具有很好的应用价值。     

Tracking of facial deformations in multi-image sequences with elimination of rigid motion of the head

The paper deals with measurement of human facial deformations from synchronized image sequences taken with multiple calibrated cameras from different viewpoints. SIFT (Scale Invariant Feature Transform) keypoints are utilized as image feature points in the first place to determine spatial and temporal correspondences between images. If no temporal match is found for an image point by keypoint matching, then the tracking of the point is switched to least squares matching provided the point has one or more spatial corresponding points in the other views of the previous frame. For this purpose, a new method based on affine multi-image least squares matching is proposed where multiple spatial and temporal template images are simultaneously matched against each search image and part of the spatial template images also change during adjustment. A new method based on analyzing temporal changes in the image coordinates of the tracked points in multiple views is then presented for detecting the 3-D points which move only rigidly between consecutive frames. These points are used to eliminate the effect of rigid motion of the head and to obtain the changes in the 3-D points and in the corresponding image points due to pure deformation of the face. The methods are thoroughly tested with three multi-image sequences of four cameras including also quite large changes of facial deformations. The test results prove that the proposed affine multi-image least squares matching yields better results than another method using only fixed templates of the previous frame. The elimination of the effect of rigid motion works well and the points where the face is deforming can be correctly detected and the true deformation estimated. A method based on a novel adaptive threshold is also proposed for automated extraction and tracking of circular targets on a moving calibration object.     

影像连接点均衡化高精度自动提取

针对空中三角测量影像连接点提取中存在的误匹配、点位多而分布不均和点位定位精度低等问题,提出了一种影像连接点均衡化高精度自动提取方法。首先采用分块SIFT技术进行特征的提取与匹配,并利用并查集数据结构进行特征点的高效多视追踪;然后采用提出的物方分块点位筛选算法对点位进行了相对均衡化的择优挑选;最后采用最小二乘匹配技术对得到的SIFT连接点坐标位置进行精化。选取中国嵩山遥感定标场的有人机影像和沙漠地区无人机影像作为试验数据,通过目视检查、像方反投影误差和检查点精度等3个指标进行了分析,结果表明本文方法有效克服了弱纹理和重复纹理导致的连接点提取与匹配困难,并改善了连接点分布的均匀性和提高了连接点的定位精度。     

基于虚拟连接点模型的机载LiDAR系统安置误差自检校

利用激光扫描直接定位的严格模型,定量分析LiDAR系统安置误差对激光脚点定位精度的影响,设计了一种安置参数自检校方法提高LiDAR数据条带间相对精度,由于离散采样的激光脚点之间不存在真实的同名连接点,提出了虚拟点模型,将虚拟连接点与真实激光脚点联系起来,并定义了两组规则从激光脚点坐标计算出连接点坐标。采用安阳市内真实LiDAR数据进行试验,证明了本文介绍的自检校方法的可行性和有效性,获得的检校参数稳定有效,可直接对原始LiDAR条带进行纠正,补偿系统安置误差。     

An algorithm for automatically computing the horizontal shift between homologous contours from DTMs

The algorithm we present here is useful for measuring the planimetric discrepancy between two models which can be represented by contours. In cartography the paradigm of such models is represented by digital terrain models (DTMs). The measure we propose is based on an area enclosed between homologous contours (two level contours, each belonging to two different models). Our measure is similar to another used previously, although, in the previous studies, the area enclosed between homologous contours was computed manually or it was restricted to particular conditions. The innovation of our approach consists of the automation of the entire process and in the constraint elimination. The main problem to solve before computing the horizontal shift measure is how to determine the area enclosed between homologous contours. This is a problem because there is no bijective correspondence between homologous contours, and therefore the identification of the homologous contours is not a trivial task. Our approach overcomes this difficulty by closing following the limit of the DTM the open contours (the contours which cut the limit of the DTM) and classifying them by the external level. We achieve the automation assisted by the Boolean polygon operations, specifically using the symmetric difference operation. Our algorithm facilitates the computation of accuracy of a DTM by comparison with another and it is applicable in such fields as hydrology (precision estimation of the hydrological features), cartographic generalization, and civil engineering. Finally we use our measure for estimating the planimetric discrepancy between the same streams derived from different DEMs (different sources and precisions).     

模糊综合评判及地形图质量的评定

本文根据模糊综合评判的原理,提出了用多级综合评判的数学模型研究地形图质量的评定问题,根据国家用图的不同需要,提出了三种具体的数学模型,並通过实例对这三种模型的实用性和有效性进行了比较,指出模型(3)是地形图质量评判的基本模型。     

Closed form solution to the twin P4P or the combined three dimensional resection-intersection problem in terms of Möbius barycentric coordinates

The twin perspective 4 point (twin P4P) problem – also called the combined three dimensional resection-intersection problem – is the problem of finding the position of a scene object from 4 correspondence points and a scene stereopair. While the perspective centers of the left and right scene image are positioned by means of a double three dimensional resection, the position of the scene object imaged on the left and right photograph is determined by a three dimensional intersection based upon given resected perspective centers. Here we present a new algorithm solving the twin P4P problem by means of M?bius barycentric coordinates. In the first algorithmic step we determine the distances between the perspective centers and the unknown intersected point by solving a linear system of equations. Typically, area elements of the left and right image build up the linear equation system. The second algorithmic step allows for the computation of the M?bius barycentric coordinates of the unknown intersected point which are thirdly converted into three dimensional object space coordinates {X,Y,Z} of the intersected point. Typically, this three-step algorithm based upon M?bius barycentric coordinates takes advantage of the primary double resection problem from which only distances from four correspondence points to the left and right perspective centre are needed. No orientation parameters and no coordinates of the left and right perspective center have to be made available. Received 1 May 1996; Accepted 13 September 1996     

A New Progressive Simplification Method for Point Cloud Using Local Entropy of Normal Angle

With the development of modern 3D measurement technologies, it becomes easy to capture dense point cloud datasets. To settle the problem of pruning the redundant points and fast reconstruction, simplification for point cloud is a necessary step during the processing. In this paper, a new method is proposed to simplify point cloud data. The kernel procedure of the method is to evaluate the importance of points based on local entropy of normal angle. After the estimation of normal vectors, the importance evaluation of points is derived based on normal angles and the theory of information entropy. The simplification proceeds and finishes by removing the least important points and updating the normal vectors and importance values progressively until user-specified reduction ratio is reached. To evaluate the accuracy of the simplification results quantitatively, an indicator is determined by calculating the mean entropy of the simplified point cloud. Furthermore, the performance of the proposed approach is illustrated with two sets of validation experiments where other three classical simplification methods are employed for contrast. The results show that the proposed method performs much better than other three methods for point cloud simplification.     

一种基于SX10点云的单株树木骨架点提取方法

骨架点是单株树木几何模型重建的基础。本文结合水平集、聚类方法和图论3种方法,采用ILMSVP和天宝SX10获取模拟与实测单株树木点云数据,并实现单株树木骨架点提取,其中模拟数据用于方法实现,实测数据用于验证。为解决单株树木点云数据缺失情况,本文重复利用水平集、聚类方法和图论方法得到较完整的单株树木骨架点。结果表明：本文采用的单株树木骨架点提取方法可行。     

Photo Orientation Under Unstable Conditions: A Robust Trial and Error Approach Using Range Ratios

Relative orientation is one of the fundamental steps in forming a photogrammetric model. Whilst relative orientation is straightforward in most cases of aerial photogrammetry this may not be the case in close range photogrammetry where the nature of the photographs and the cameras being used often raise unfamiliar problems. This paper addresses the problem of performing relative orientation of non-metric photographs under unstable conditions where there are no ground control points and the ray intersection geometry is poor. The relative orientation problem is approached from several viewpoints, including a method that relies on little information about the exterior or relative orientation parameters. The solution is based on an iterative process, in which a set of three virtual control points (VCPs) is created, using photo coordinates and range ratios. The validity of the VCPs is checked during every iteration by an estimation of the coplanarity condition using two or more conjugate checkpoints. During the iteration a database is formed, which is then used to obtain a net of VCPs that fulfil the coplanarity condition for the checkpoints. The VCPs are then used to compute the "relative" exterior orientation by space resection. The proposed method was tested on both simulated and real photographs and was found to have some advantages over more commonly used techniques.