基于等高线空间关系的鞍部点提取方法

鞍部点是反映地表形态起伏变化的重要地形特征点之一,准确地提取鞍部点有利于地形的空间关系和结构特征分析。现有的鞍部点提取方法通常是直接基于规则格网DEM数据,无法顾及鞍部点与周围地形的空间拓扑关系和复杂地形对其的影响,不仅产生大量的伪鞍部点,而且忽略一些关键地区的鞍部点。本文根据鞍部点的地形形态特征,设计了一种基于等高线数据的鞍部点提取算法。该算法利用等高线闭合的特征,将等高线按照一定规则转成等高面数据,再利用等高面之间的相邻拓扑关系实现递归查找并自动提取鞍部点。实验结果显示:①鞍部点的数量和位置与等高距的大小显著相关,在一定尺度范围内,等高距越小,提取出鞍部点越多,位置精度也逐渐提高;②与基于规则格网DEM数据提取方法相比,该方法能更有效的过滤大量伪鞍部点,提高了鞍部点的提取精度,同时也降低了鞍部点提取算法的复杂度;与基于等高线的增量缓冲方法(Incremental Buffering Algorithm)相比,本文的方法能有效提高鞍部点提取的完整性,更适用于本文DEM的尺度即5 m DEM数据。     

面向点云数据的黄土丘陵沟壑区沟沿线自动提取方法

基于点云数据地表建模中,沟沿线既是点云去噪的重要分割线,又是黄土丘陵沟壑区最典型的地形特征线。因此,本文提出了面向点云数据的基于多尺度格网采样和坡度分割的沟沿线自动提取方法,即在适宜格网尺度采样的基础上构建地表模型,利用沟沿线周围坡度陡变的特性提取缓坡面与陡坡面的分界线,从而生成沟沿线。通过对8个样区反复实验,确定了各自的最佳格网尺度,并发现格网尺度随点密度的增大而迅速减小,而后趋近平稳的幂函数关系,同时随机选取了另外5个样区验证了该关系的适用性。最后,通过全域分块计算得到了该地区的完整沟沿线。相比于人工识别的沟沿线,本文方法提取沟沿线的精度在0.5 m缓冲范围内为85%,效果较好,且位置更为精确。产生差异的原因主要是面向点云数据的沟沿线有更多的细节,沟沿线曲率更大,造成了其长度明显大于手动提取结果。该方法有助于提高丘陵沟壑区点云数据去植被处理和地形表面重建精度。     

数字高程模型精度评定的基本理论

在剖析随机误差与逼近误差区别的基础上 ,指出数字高程模型 (DEM)内插误差属于逼近误差 ,精度评价方法不应继续延用“中误差”概念 ,而应采用“逼近误差”,给出了常用 DEM内插方法的误差评定模型 ,分析了等高线间距对 DEM内插精度影响的作用机制 ,最后解释了 DEM精度分析中用“中误差”概念无法解释的几个问题。     

Uncertainty of slope length derived from digital elevation models of the Loess Plateau,China

Although many studies have investigated slope gradient uncertainty derived from Digital Elevation Models（DEMs）, the research concerning slope length uncertainty is far from mature. This discrepancy affects the availability and accuracy of soil erosion as well as hydrological modeling. This study investigates the formation and distribution of existing errors and uncertainties in slope length derivation based on 5-m resolution DEMs of the Loess Plateau in the middle of China. The slope length accuracy in three different landform areas is examined to analyse algorithm effects. The experiments indicate that the accuracy of the flat test area is lower than that of the rougher areas. The value from the specific contributing area（SCA） method is greater than the cumulative slope length（CSL）, and the differences between these two methods arise from the shape of the upslope area. The variation of mean slope length derived from various DEM resolutions and landforms. The slope length accuracy decreases with increasing grid size and terrain complexity at the six test sites. A regression model is built to express the relationship of mean slope length with DEM resolution less than 85 m and terrain complexity represented by gully density. The results support the understanding of the slope length accuracy, thereby aiding in the effective evaluation of the modeling effect of surface process.     

规则格网DEM局地坡面凸凹性精度分析

DEM地表形态精度分析理论与方法的建立,对DEM数据的生产和广泛应用具有重要意义。本文从局地坡面形态的凸凹性角度,剖析规则格网DEM格网点位置、格网分辨率对DEM局地坡面凸凹性的影响,以期进一步完善和发展DEM质量分析的理论与方法。论文首先阐述了DEM局地坡面凸凹性的基本概念,研究建立了规则格网DEM的局地坡面凸凹性量化分析方法,并以黄土丘陵5、10、15、25、……、155 m DEM为例,采用比较分析方法研究了局地坡面凸凹性随DEM格网点位置和格网分辨率的变化特征。研究表明:对于本研究中的1:5万DEM,10 m(跃变率≤0.3%)是其最佳的格网分辨率阈值,当DEM实际格网分辨率高于该阈值时,实际DEM与最佳格网分辨率DEM具有近乎相同的局地坡面凸凹性,主要在正地形与负地形的过渡区域会发生不同程度的坡面凸凹性变化;当DEM实际格网分辨率低于该阈值时,实际DEM的局地坡面凸凹性,会随着DEM格网点布设位置和DEM格网分辨率发生较大的不确定性变化。     

基于SWAT模型和多源DEM数据的流域水系提取精度分析

流域水系是研究水文水资源、地貌演化和生态环境及水土治理等的基础数据,高精度的水系提取对流域研究十分重要。本文以空间分辨率均为30 m的 AW3D30 DSM、SRTM1 DEM和ASTER GDEM2数字高程模型作为基本的地形数据,基于SWAT模型提取犟河流域水系,通过河网“套合差”、水系相对误差、Google Map水文数据及蓝线河网对提取结果进行误差分析与综合评价,探讨河道剖面和地形特征对水系提取精度的影响。结果表明：① 集水面积阈值是决定河网水系提取精度的关键参数,阈值越大,提取的河网密度越小,反之提取的河网密度越大;② 基于河网密度与集水阈值二阶导数的幂函数与直线相切的数学求值方法确定流域最佳集水面积阈值,能避免最佳集水阈值取值的主观性,提取的河网水系与实际河道相符;③ AW3D30 DSM数据提取的流域河网水系与Google Map高分辨率影像的水系偏差最小,且AW3D30 DSM数据提取的水系与蓝线河网的河网“套合差”和水系相对误差值均最低,能真实反映中低山丘陵山区流域水系发育的疏密程度,吻合度最好;④ 多源DEM数据提取结果均显示为河床比降大和横剖面曲线为窄深式的“V”形河谷提取的水系精度高于河床比降小和横剖面曲线为 “碟”形河谷的提取精度;⑤ AW3D30 DSM数据的地形起伏和坡度标准差最大,有利于山区河网水系的提取。因此,基于SWAT模型和AW3D30 DSM数据提取的山区流域水系可最大限度反映流域水系的真实情况,精度最高,此方法和数据源可应用于中低山丘陵山区流域的水系提取研究。     

基于改进主动轮廓模型的无人机影像矿区地裂缝提取

矿区地裂缝精准识别对防灾、减灾和生态环境修复具有重要意义。针对高分辨率无人机影像较难自动精确提取地裂缝的问题,本文提出了一种基于改进主动轮廓模型的无人机影像矿区地裂缝提取方法。首先,采用Otsu算法计算背景和地裂缝初值作为先验知识;其次,构建背景和地裂缝初值的提取能量函数,并引入到传统CV主动轮廓模型,增强地裂缝提取的针对性;最后,通过轮廓的不断演化实现地裂缝的提取。以内蒙古扎赉诺尔矿区为研究区、无人机影像为数据源,采用改进主动轮廓模型方法进行地裂缝提取,并与传统的Canny边缘检测算法、支持向量机（SVM）、最大似然（MLM）和传统CV主动轮廓模型方法进行对比分析。结果表明：在地物类型较为单一的小范围区域,传统的Canny边缘检测算法和传统CV主动轮廓模型提取效果较差,改进主动轮廓模型、SVM和MLM共3种方法均可以取得较好的效果,其中,改进主动轮廓模型方法精度最高;在地物类型相对复杂的大范围区域,传统的Canny边缘检测算法、SVM、MLM和传统CV主动轮廓模型方法存在较多的漏提和误提,Kappa系数均低于0.7,而本文改进主动轮廓方法依然可以取得较好的效果,Kappa系数达到0.9左右。因此,本文提出的方法通过引入先验知识可有效提高地裂缝提取的精度和稳定性。     

基于DEM的黄土地貌逼近度因子构建及特征分析

沟谷源点作为沟沿线上最为活跃的部位,其分别到上游分水线、下游沟谷线的流线空间比对关系,是表征三线空间结构的重要突破口。它在空间上向流域分水线逼近的程度,是量化黄土流域地貌系统发育程度的重要切入点。为探究黄土地貌区流域沟谷源点向流域分水线逼近的程度,揭示黄土流域地貌发育进程及该进程所表现的主要侵蚀方式,本文从水平和垂直2个维度,构建量化三线空间结构关系的核心因子--逼近度(PI),其中包括水平逼近度(HPI)和垂直逼近度(VPI),基于5 m分辨率数字高程模型,在陕北黄土高原遴选了包含16种地貌类型的42个样区,利用数字地形分析方法,探讨其平均值(MHPI, MVPI)的空间分异规律。另外,选择其中南北序列分别代表黄土塬、残塬、梁状丘陵沟壑和峁状丘陵沟壑区的淳化、宜君、甘泉和绥德4个地区,完备包含一至五级别沟谷的典型流域为重点实验样区,探讨陕北黄土高原流域尺度平均逼近度变异指数(MPIV)序列性特征。实验结果表明:①陕北黄土高原平均逼近度(MPI)存在强烈的空间自相关,MHPI在南北序列上先增大后减小,东西方向上逐渐减小,在黄河沿岸峡谷丘陵区MHPI达到最大;MVPI由西南-东北先减小后增加,由西北-东南逐渐减小,在渭北黄土台塬区达到最小;②在流域尺度上,MPIV值的正负与黄土塬区、丘陵沟壑区敏感相关;③4个重点样区MHPI、MVPI与其它地形因子存在南北序列上的一致性。104个外部汇流区平均水平逼近度与平均坡度相关性较好(P=0.43, a<0.001),平均垂直逼近度与面积高程积分强烈相关(P=0.75, a<0.001)。平均逼近度指标综合考量了黄土高原地区最典型的三条具有结构控制意义特征线的空间关系,对黄土地貌的发育程度有明显的指示性意义。     

融合网格注意力阀门和特征金字塔结构的高分辨率遥感影像建筑物提取

在高分辨率遥感影像中提取建筑物轮廓是地区基础建设信息统计的一项重要任务。适应性较强的深度学习方法已在建筑物提取研究中取得较大进展,受网络模型对影像特征表达的局限性,存在局部建筑轮廓边缘模糊的问题。本研究提出一种基于注意力的U型特征金字塔网络（AFP-Net）可以聚焦高分遥感影像中不同形态的建筑物结构,实现建筑物轮廓的高效提取。AFP-Net模型通过基于网格的注意力阀门Attention Gates模块抑制输入影像中的无关区域,凸出影像中建筑物的显性特征;通过特征金字塔注意力Feature Pyramid Attention模块增加高维特征图的感受野,减少采样中的细节损失。基于WHU建筑物数据集训练优化AFP-Net模型,测试结果表明AFP-Net模型能够较清晰地识别出建筑物轮廓,在预测性能上有更好的目视效果,在测试结果的总体精度和交并比上较U-Net模型分别提高0.67%和1.34%。结果表明,AFP-Net模型实现了高分遥感影像中建筑物提取的结果精度及预测性能的有效提升。     

A new strategy combined HASM and classical interpolation methods for DEM construction in areas without sufficient terrain data

Spatial interpolation is an important method in the process of DEM construction. However, DEMs constructed by interpolation methods may induce serious distortion of surface morphology in areas lack of terrain data. In order to solve this problem, this paper proposes a strategy combining high-accuracy surface modeling(HASM) and classical interpolation methods to construct DEM. Firstly, a triangulated irregular network(TIN) is built based on the original terrain data, and the area of the triangles in the TIN is used to determine whether to add supplementary altimetric points(SA-Points). Then, classical interpolation methods, such as Inverse Distance Weighted(IDW) method, Kriging, and Spline, are applied to assign elevation values to the SA-Points. Finally, the SA-Points are merged with the original terrain data, and HASM is used to construct DEM. In this research, two test areas which are located in Nanjing suburb in Jiangsu Province and Guiyang suburb in Guizhou Province are selected to verify the feasibility of the new strategy. The study results show that:(1) The combination of HASM and classical interpolation methods can significantly improve the elevation accuracy of DEMs compared with DEM constructed by a single method.(2) The process of adding SA-Points proposed in this study can be repeated in many times. For the test areas in this paper, compared with the results with only one execution, the results with more executions are in much more accordance with the actual terrain.(3) Among all the methods discussed in this paper, the one combined HASM and Kriging produce the best result. Compared with the HASM alone, absolute mean error(MAE) and root mean square error(RMSE) of the best result were reduced from 1.29 m and 1.83 m to 0.68 m and 0.45 m(the first test area), and from 0.32 m and 0.38 m to 0.21 m and 0.28 m( The second test area).     

黄土高原地形因子间关联性的神经网络分析

不同的地形因子从不同侧面反映地面的起伏特征或空间变异,各因子之间所存在的相互关联、相互制约、相互影响的特征,在很大程度上揭示了地形发育与空间变异的内在本质。本文以黄土高原丘陵沟壑区的1∶10000和1∶50000两种比例尺的15个实验样区为样本,应用BP神经网络模型,探讨不同比例尺的地形定量因子与地面坡度之间的关联特征及其变化规律。实验结果表明:①利用神经网络分析方法可以有效定量评价地形因子间的关联性;②该研究方法有助于地学分析中DEM尺度的选择,地形因子的确定及其相关关系的量化。     

太阳辐射模拟的DEM尺度影响因素分析

利用DEM分析方法为起伏地形下太阳辐射模拟提供了条件。然而DEM的尺度效应对太阳辐射模拟产生很大的影响。本文以黄土丘陵和秦岭高山为实验样区,以一系列不同分辨率的DEM,模拟日、月及年总天文辐射,比较其误差大小及地形分析的差异。实验结果显示,随着DEM栅格分辨率的降低,天文辐射逐渐增大;当栅格分辨率大于地面相对高差时,模拟的天文辐射基本稳定,坡向、坡度等地形因子对辐射的再分配作用已不显著。黄土丘陵区,当DEM分辨率接近1000m时,天文辐射年总量的相对误差为22%,超过该区域一个冬季的天文辐射量。秦岭山区地表起伏大,误差较黄土丘陵地区小,最大相对误差为17.8%。总之,地表起伏越小,DEM栅格分辨率对太阳辐射模拟的影响越大。     

A DEM fusion method for ascending and descending orbit StereoSAR DEMs in complicated mountain areas considering SAR echo intensity

To extract the high-quality DEM in complicated mountain areas,a DEM fusion method for ascending and descending orbit StereoSAR DEMs considering Synthetic Aperture Radar(SAR)echo intensity is proposed.After the analysis for the influence of terrain features and SAR side-looking imaging characteristics on radar echo intensity and DEM accuracy,four Terras AR-X images with the stripmap mode and the 3 m spatial resolution covering a certain area of Maoxian County,Sichuan Province,China,was selected as the experimental area.StereoSAR technology was used to extract the ascending orbit StereoSAR DEM and the descending orbit StereoSAR DEM,respectively,and the corresponding radar echo intensity map was calculated.Then,while comparing the radar echo intensity corresponding to the same point position,DEM fusion was carried out,and the accuracy of DEM before and after the fusion was analyzed with the ground points measured by GNSS-RTK as reference data.Finally,a high-quality DEM with a 3 m spatial resolution in the experimental area was obtained.The DEM accuracy was improved on all slopes,and the mean absolute deviation(MAD)improved to 4.798 m,the standard deviation(SD)improved to 6.087 m and the LE90 improved to 40.48 m.The experimental results indicate that the fusion method of highresolution ascending and descending orbit StereoSAR DEMs considering SAR echo intensity can effectively extract DEM with high accuracy and reliability,which can provide technical support for obtaining highquality terrain information in similar areas.     

基于DEM的沟谷特征点提取与分析

沟谷特征点是反映沟谷地貌空间形态分布的重要点位,也是研究沟谷地貌演化过程与机理的关键要素。因此,对不同沟谷特征点的有效提取,是沟谷形态研究的重要基础。本文采用DEM及其在水文分析中的多种衍生数据,通过流向追踪、邻域特征判断等一系列方法,实现对径流节点、径流源点、汇流源点、潜在裂点与流域出口点的快速、准确提取。同时,对沟谷特征点进行有效分级,本文基于Strahler河流分级法建立了相应的分类标准,对沟谷特征点进行了自动分类。通过使用陕西省宜君典型样区5m分辨率的DEM数据进行实验,发现新算法计算效率高,结果准确,对潜在裂点也进行了有效探测,验证了算法的有效性。最后,对特征点提取的数据影响进行了详细分析。     

于DEM光照晕渲模拟的梯田自动提取方法

梯田作为黄土高原最为典型的人工地貌之一,有重要的农业生产和水土保持价值。传统的梯田自动提取仅限于梯田所在区域的范围划定,未能对梯田田坎线实现有效的自动化提取。鉴于此,本文提出了一种基于光照晕渲模拟的梯田快速提取方法。首先,对无人机航测生成的1 m分辨率的数字高程模型（DEM）进行4个方向的光照晕渲模拟,并相加取平均值;然后,通过适当的阈值对均值图像进行二值化,并掩膜掉沟谷等非梯田区域;最后,基于二值化图像自动矢量化得到梯田田坎线,并通过适当的长度阈值进行碎线过滤提高提取精度。本文以陕西省长武县王东沟流域为实验样区进行了实验,结果表明,该方法提取结果的准确率为89.09%,具有较好的提取精度。此外,对该方法涉及到的参数进行了讨论,表明光照模拟的方向角采用2个正交的对称方向对、高度角采用田坎坡度的反正切值、二值化阈值采用t=180-σ的经验公式,可以满足黄土高原的梯田自动快速提取。     

Comparison of DEM accuracies generated from different stereo pairs over a plateau mountainous area

Digital elevation models(DEMs) can be quickly and conveniently generated using very high resolution(VHR) satellite stereo images. Previous studies have evaluated and compared DEM accuracy based on VHR satellite stereo pairs collected by different satellite sensors. However, few studies analyzed the accuracy of a DEM based on stereo image pairs from a satellite with the same orbit and different orbits for a region with significant topographic fluctuations in the plateau area. Referring to former studies, this paper had two objectives: to generate a digital elevation model(DEM) and evaluate its horizontal and vertical accuracy over a plateau area with high relief; and to study the mapping capability of multiorbit and multitemporal stereo pair images in the plateau mountainous region. To achieve these objectives, we collected the 2015 Worldview-2 stereo image pair and another three World View-2 images acquired in 2013, 2014, and 2015. First, the 2015 DEM was obtained using a strict physical model based on along-track stereo image pairs, and the reliability of the DEM was verified with field data. Then, the images obtained in 2013, 2014, and 2015 were combined into different-orbit stereo image pairs, DEMs were produced using rational function models, and the DEMs were verified using field data and the 2015 DEM as standards. The results showed that the relief degree has a particular influence on the DEM, and the precision of the DEM decreases as the topographic relief increases. Off-nadir angles can also influence DEM accuracy, with a larger angle corresponding to a lower DEM accuracy. The research also shows that the DEM obtained from four sets of experiments meets the accuracy requirement of a 1:5,000 digital elevation map, digital line graphic(DLG), and digital orthophoto map(DOM). Among these four groups of DEMs, the one based on the 2015 stereo pairs with the same satellite achieved the highest precision.     

Arc Hydro Tools及多源DEM提取河网与精度分析——以洮儿河流域为例

本文简要分析了HYDRO1K、SRTM3-2、ASTER GDEM 3种数据源基本特征,对数据的水平空间分辨率、现势性、覆盖范围、误差来源进行了比较;阐述了Arc Hydro Tools提取数字河网、划分流域及子流域方法的关键步骤;并以洮儿河流域为研究对象,分析了有(无)河网辅助条件下,3种初始DEM在不同地貌类型中数...     

A simple method to extract glacier length based on Digital Elevation Model and glacier boundaries for simple basin type glacier

Glacier length is a key morphological element that has many glaciological applications; however, it is often difficult to determine, especially for glaciers that cover larger spatial areas or those that exhibit frequent temporal change. In this paper, we describe a new ArcGIS-based method that can derive glacier flow lines for determining glacier length based on digital elevation model and glacier outlines. This method involves (1) extraction of the highest and lowest points on a glacier, (2) calculation of 10-m contour lines on the glacier from 10 m to 100 m height, and (3) connection of the midpoints of each contour line with the highest and the lowest points in order to create a flow line, which is subsequently smoothed. In order to assess the reliability of this method, we tested the algorithm’s results against flow lines calculated using field measurements, analysing data from the Chinese Glacier Inventory, and manual interpretation. These data showed that the new automated method is effective in deriving glacier flow lines when contour lines are relatively large; in particular, when they are between 70 m and 100 m. Nonetheless, a key limitation of the algorithm is the requirement to automatically delete repeated and closed curves in the pre-treatment processes. In addition to calculating glacier flow lines for derivation of glacier length, this method also can be used to effectively determine glacier terminus change.     

基于地貌特征的数字高程模型融合方法

数字高程模型（Digital Elevation Model, DEM）是一种至关重要的空间信息,广泛应用于各行各业。其中,ASTER GDEM与SRTM几乎覆盖了全球陆域,为地学研究提供了非常实用的高程数据支撑,但是由于二者传感器采集数据原理的不同,使得高程数据在不同地貌条件下的高程精度亦存在程度不一的误差。本文提出了一种新型的基于地貌特征的DEM融合方法,使得融合GDEM与SRTM后的DEM数据,消除了地貌特征的影响、显著地提高了DEM质量。该方法主要分为地理配准和高程融合2个步骤：①基于河流线对等线性地貌特征的位置数据,构建了GDEM与SRTM的水平偏移相关的误差评价函数,采用多级网格搜索法求得DEM间的水平偏移距离,实现对DEM的配准;②按照DEM高程值在不同地貌单元及边界线附近的高程变化特征,建立地貌分区的高程融合模型来融合两种地理配准后的DEM高程,尤其是实现了地貌单元边界线附近的高程平滑过渡。本文以怀柔北部地区为实验区,以1:5万地形图为参考,对2种DEM数据进行融合,统计结果表明：① 融合DEM在各地貌单元的误差均显著下降,地形表达较之融合前更加精确;② 高程差呈现正态分布,明显区别于融合前DEM不对称的多峰分布形态,说明地貌影响被有效地剔除;③ GDEM和SRTM数据的精度对坡度有较大依赖性,融合后DEM的精度在不同坡度范围下均优于GDEM和SRTM,显著降低了融合前DEM对坡度的依赖程度;④ 在不同坡向下,GDEM和SRTM的RMSE取值波动较大,融合DEM的RMSE取值在各方向表现稳定,高程精度较GDEM和SRTM有显著提高。     

基于DEM的5节点二阶差分坡度算法模型与实验分析

 坡度作为最基本的地形因子,是构建地学分析模型的基础数据,基于格网DEM的坡度计算模型的精度分析一直是地学分析的重要研究内容。本文通过分析计算坡度的 差分模型误差来源及特征,提出了一种新的模型——5节点二阶差分模型。其基本思想是顾及多距离邻近点高程信息,采用5×5分析窗口,建立单倍格网和两倍格网的差分计算线性组合模型。并通过数学分析的理论推导证明该模型可以显著提高计算精度。实验选取典型数学曲面并对其采用不同的分辨率(1m和5m)进行离散化精度验证。计算结果表明,与现有的差分模型相比,5节点二阶差分模型可以显著提高坡度的精度,对应于两种分辨率的DEM,该模型计算坡度的精度可以分别提高7×10⁴倍、3×10³倍以上,表明该模型对于高分辨率DEM数据更为适用。本研究不仅丰富了数字地形分析的方法体系,为相关地学模型分析提供更精确的坡度数据,而且为将来改进地形曲率因子等计算模型提供了一定的参 考和借鉴。