EN中文

地球信息科学学报 >

2018 , Vol. 20 >Issue 3: 281 - 290

DOI: https://doi.org/10.12082/dqxxkx.2018.170350

地球信息科学理论与方法

顾及多分析尺度的地形部位面向对象分类方法

  • 江岭 , 1, 2, * ,
  • 凌德泉 3 ,
  • 赵明伟 1, 2 ,
  • 王春 1, 2 ,
  • 曾微波 1, 2
展开
  • 1. 滁州学院 安徽地理信息集成应用协同创新中心,滁州 239000
  • 2. 安徽省地理信息智能感知与服务工程实验室,滁州 239000
  • 3. 南京信息工程大学 地理与遥感学院,南京 210044

作者简介：江 岭(1987-),男,安徽人,博士,副教授,研究方向为数字地形建模及高性能地学计算。E-mail:

收稿日期: 2017-07-28

  要求修回日期: 2017-10-30

  网络出版日期: 2018-03-20

基金资助

国家自然科学基金项目（41501445、41571398）

安徽省自然科学基金项目（1608085QD77）

安徽省高等学校自然科学研究项目（KJ2015A171）

收起

Object-oriented Terrain Position Classification Based on Multi-scale Geomorphons

  • JIANG Ling , 1, 2, * ,
  • LING Dequan 3 ,
  • ZHAO Mingwei 1, 2 ,
  • WANG Chun 1, 2 ,
  • ZENG Weibo 1, 2
Expand
  • 1. Anhui Center for Collaborative Innovation in Geographical Information Integration and Application, Chuzhou University, Chuzhou 239000, China
  • 2. Anhui Engineering Laboratory of Geo-information Smart Sensing and Services, Chuzhou University, Chuzhou 239000, China
  • 3. School of Geography and Remote Sensing, Nanjing University of Information Science & Technology, Nanjing 210044, China;
*Corresponding author: JIANG Ling, E-mail:

Received date: 2017-07-28

  Request revised date: 2017-10-30

  Online published: 2018-03-20

Supported by

Program of National Natural Science Foundation of China, No.41501445, 41571398

Program of Provincial Natural Science Foundation of Anhui, No.1608085QD77

Key Project of Natural Science Research of Anhui Provincial Department of Education, No.KJ2015A171.

Copyright

《地球信息科学学报》编辑部 所有
Fold

摘要

地形部位是地表形态的基本单元,其分类和提取在地貌发育、数字土壤制图、景观生态制图等领域有着重要的应用。康鑫等提出的多尺度Geomorphons地形部位分类法（简称多尺度Geomorphons法）利用高程相对差异信息和地形部位多尺度特征,可避免受地形属性计算及单一分析尺度约束而误分类,然而其存在分类破碎及分析尺度域难以确定的问题。基于此,本文以多尺度Geomorphons法为基础,提出了其适宜分析尺度域确定方法,建立了以初始地形部位数据层组合的对象多尺度分割和分类方法,进而构建了顾及多分析尺度的地形部位面向对象分类方法。以陕北黄土高原区域5 m分辨率DEM为实验数据,对面向对象分类方法进行了验证与评价。实验结果表明：①均值变点法可有效解决分析尺度域难以确定的问题,实验样区适宜分析尺度域为[5×5, 33×33]栅格单元;②以0,255为二值化的地形部位数据层组合适用于多尺度分割,尺度、形状及紧致度参数组合影响分割结果,且对于实验样区存在最优分割参数;③与多尺度Geomorphons法相比,本文方法得到的地形部位分类结果完整性较好,在地表形态对应和地理认知等方面更具合理性。

关键词： 地形部位; DEM; 面向对象; 多尺度分析; 数字地形分析

本文引用格式

江岭 , 凌德泉 , 赵明伟 , 王春 , 曾微波 . 顾及多分析尺度的地形部位面向对象分类方法[J]. 地球信息科学学报, 2018 , 20(3) : 281 -290 . DOI: 10.12082/dqxxkx.2018.170350

Abstract

Terrain position is the basic morphologic feature on the surface of the Earth. The classification and extraction of terrain position have been widely applied in many research fields such as landform evolution, digital soil mapping and landscape ecological mapping. Proposed by Kang X et al. (2016), the multi-scale Geomorphons method maps terrain position by recognizing the morphology of each interest cell in a DEM according to its relative altitudes within the neighboring window. Multi-scale Geomorphons method can avoid the shortnesses of other classificaton methods, which are caused by different terrain attributes and a single analysis scale. However, there are still some drawbacks in the multi-scale Geomorphons method. For example, the classification results are fragmented and the domain of the analysis scale is difficult to determine. To solve the above problems, this paper presents a new method to classify terrain position, which is based on object-oriented segmentation and multi-scale Geomorphons. First of all, we propose an approach of determining the domain of optimal analysis scale of the multi-scale Geomorphons method. Then, the multi-scale segmentation and classification methods are constructed according to the initial terrain position data via the multi-scale Geomorphons method. At last, the presented method is evaluated by the experimental data of the DEM with 5 m resolution in the loess plateau region of northern Shaanxi. The experimental results show that: (1) the method of mean change-point analysis can effectively solve the problem which is difficult to determine the domain of the analysis scale of the multi-scale Geomorphons method. The domain of optimal analysis scale of the sample area is 5×5 to 33×33 cells. (2) The layer of each terrain position type with the value 0 for non-type cells and 255 for type cells is suitable for multi-scale segmentation. The parameters (i.e. scale, weight of shape and weight of compactness) for multi-scale segmentation have deep influence on segmentation results. There is optimal segmentation parameters for a experimental region. There is optimal segmentation parameter for an experimental region. (3) Comparing with the multi-scale Geomorphons method, the classification results of the present approach are more integrity and reasonable in the aspects of morphology correspondence and geological interpretation.

Key words： terrain position; DEM; object-oriented; multi-scale analysis; DTA

1 引言

作为地表形态的基本单元,地形部位是指在一定尺度下由地形属性特征相似且空间邻近的单元组合形成的区域[1,2]。不同的地形部位组合构成了形态多样的地形坡面,进而形成了特征迥异的地貌类型。通常土壤、水文以及植被等地理过程随着地形部位的变化而变化,甚至突变。因此,地形部位分类和提取是研究地貌发育、土壤侵蚀和调查、景观生态分类与制图,以及分布水文模型等方面的基础[3,4,5,6,7]。
目前,地形部位分类方法多样,体系尚不统一。地形部位分类方法可以归纳为以下3种：聚类法、原型法和规则知识法[8]。聚类法是通过聚类的思想对地形属性集合进行聚类,再将聚类出的类型识别为地形部位[9,10,11]。聚类法虽易于操作,但结果易受地形属性选择所影响,且部分地形部位难以进行地学解释[12]。原型法通过设置原型部位,并计算目标位置与原型部位间的相似度,从而实现地形部位分类[13,14,15]。该类方法虽能顾及地形部位间的渐变特征,但在地形属性选择及模糊推理方面存在主观性,且挖掘分类知识过程计算量大。规则知识法根据区域特征和专家知识,显式表达和量化规则化地形部位,通过明确多个地形因子的取值范围进行地形部位划分[2,16-17]。相比前2种方法,规则知识法可得到地学意义明确的结果且计算简单。相对于存在量化规则难的基于多地形属性规则知识法而言,规则知识法中的相对高差法直接根据待分类空间位置与其邻域范围内的相对高程差异来判定待分类类型,可有效地避免因地形属性集选择而引起的误分类以及分类不确定等问题[2,17-18]。
作为相对高差法的典型代表,多分析尺度综合的Geomorphons方法（简称多尺度Geomorphons法）对Geomorphons方法进行了优化,有效顾及了地形部位多尺度特征,分类结果更为合理[2]。然而,多尺度Geomorphons法本质上隶属于逐栅格单元分类方法,其分类结果较为破碎,“椒盐”现象显著。与此同时,多尺度Geomorphons法还存在分析尺度域难以确定的问题,不同的分析尺度域对分类的影响较大。为此,本文拟引入面向对象分类思想,构建顾及多分析尺度的地形部位面向对象分类方法,以解决多尺度Geomorphons法所存在的不足。

2 多尺度Geomorphons地形部位分类

Geomorphons方法基于利用地形开放度指数计算的待分类栅格单元与8个方向邻域之间高、低、相等3种关系,组合出38种局域几何形态,再根据地学含义进行归整,进而依据所设计的映射表将其对应到10种地形部位[17]（图1）。
Fig. 1 Diagram of terrain position

图1 地形部位示意图[17]

在Geomorphons方法的基础上,康鑫等[2]提出了多尺度Geomorphons法,采用Douglas-Peucker线简化算法替代原方法中的开放度算法,并通过综合多分析尺度（即分析尺度域）下分类结果得出地形部位分布情况。多尺度Geomorphons法避免了原方法中因单一尺度而引起的分类结果与地形部位所应具有的多尺度特征不相适应问题。同时,多尺度Geomorphons法引入熵值以度量分类结果的不确定性,且熵值大小也可表征不同分析尺度域下分类结果的差异性。栅格单元（i, j）的熵值计算公式为：
En tr op y i , j = - ∑ x = 1 10 m ij ( x ) lo g 2 m ij ( x ) lo g 2 10 其中 , m ij x = ∑ k = 1 n m ij k ( x ) / k （1）
式中： m ij k （ x ） 为栅格单元（i, j）在尺度k上对第 x（x=1,…,10）类地形部位的隶属度;mi,j（x）为整个分析尺度域（1,…,n）内对x类地形部位的隶属度。

3 面向对象地形部位分类

对于多尺度Geomorphons法而言,分析尺度域的选择因使用者经验而存在较大主观性,不同的分析尺度域对分类结果存在无法规避的影响。与此同时,多尺度Geomorphons法还存在逐栅格单元分类方法难以避免的分类结果破碎问题,这不仅与地形部位定义特征不相适应,而且在地学解释上也不具合理性。为此,本文构建了顾及多分析尺度的地形部位面向对象分类方法（图2）：首先,采用均值变点法确定分析尺度域,根据多尺度Geomorphons法得到初始分类结果;其次,设定地形部位数据分层规则,获取地形部位数据层组合,并分析确定最优分割参数,进而得出分割结果;最后,基于分割对象,根据制定的分类规则及不确定性计算方法,获得地形部位分类结果及其不确定性分布。
Fig. 2 Flowchart of the presented method

图2 技术流程图

3.1 分析尺度域确定

对多尺度Geomorphons法而言,分析尺度域确定是其首要条件。分析尺度域的变化会引起分类结果的显著差异,进而产生分类结果熵值的变化。相关研究表明,随着分析尺度域的扩大,地形部位多尺度特征越明显,分类结果熵值不断增加,平均熵值变化率不断减小,且变率由陡变缓,呈对数曲线分布。这就意味着,平均熵值变化曲线必定存在一个变率由大变小（即曲线由陡变缓）的点,且该变点是唯一的,适宜分析尺度域即为该变点所对应的分析尺度范围。在诸多变点分析方法中,均值变点法是检验变点最为有效的方法之一,已在众多地学研究中广泛应用[19,20]。本文采用均值变点法来寻找适宜分析尺度域,其计算过程如下：设起始分析尺度（分析窗口）为W1：5×5栅格单元,依次窗口步距增加2得到Wn（n为增加的次数）。对于序列分析尺度域[W1,Wn],得到地形部位平均熵值样本序列H0：
（1）令 i = 2,3 , 4 , ⋯ , N ,对每个i将样本分为 2段： X 1 , X 2 , ⋯ , X i - 1 和 X i , X i + 1 , ⋯ , X n 。计算每段样本的 X i 1 ¯ 和 X i 2 ¯ 及统计量。
S i = ∑ t = 1 i - 1 ( X t - X i 1 ¯ ) 2 + ( X t - X i 2 ¯ ) 2 （2）
（2）计算总体样本的统计量。
X ̅ = ∑ t = 1 N X t N （3）
S = ∑ t = 1 N X t - X ¯ 2 （4）
均值变点法的原理在于变点的存在会使原样本的统计量S与样本分段之后的统计量Si之间的差距增大,S与Si差值最大的点即为所求的变点。在分析出适宜分析尺度域后即可采用多尺度Geomorphons法进行地形部位分类。值得说明的是,根据平坡的坡度一般小于3°[21],本文将多尺度Geomorphons法中角度参数设为3°。

3.2 面向对象分割

作为当前处理高精度“影像”数据的主流方法,面向对象分析法依据栅格单元异质性相近程度将 “影像”分割为多个对象,并将对象作为分析单元进行后续分析[11,22]。在进行分割时,多波段或单波段“影像”数据是基础。本文以多尺度Geomorphons法初始分类结果组合作为“影像”数据。由于初始分类结果中类型值（1-10）无数值含义,而在分割时,异质性是通过计算对象组内方差得到,故不能直接将分类结果作为“影像”。充分借鉴影像灰度级别,且最大程度消除不同地形部位之间数值差异影响,本文提出了由类型值空间数据生成多波段“影像”的变换分层规则：将每一类地形部位分为一个数据层（即共有10个数据层）,并将每个数据层地形部位所在栅格单元值设置为255,其余栅格单元值设置为0。
在分割过程中,选择适宜的分割尺度是关键。过大的分割尺度会使较小的对象丢失,而过小的分割尺度又会导致过分割现象,使得分割对象过于破碎。除分割尺度外,分割参数还包括有形状（与光谱和为1）、紧致度（与平滑度和为1）。分割参数对于对象分割有重要的影响。最优分割参数表现为目标地物与分割对象的边界基本吻合,分割对象不会太破碎,边界较为清晰。对于最优分割尺度确定而言,应用最为广泛的方法为局部方差变率法,即通过统计分析不同分割尺度下对象的方差和方差变率探测合适的分割尺度[23]。本文以形状和紧致度步长为0.2的变量,对二者每一参数组合,运用局部方差变率法探测最优分割尺度,并将三者组合作为分割参数候选。对于每一候选分割参数的对象,通过叠加地形判定对比分析,得出较为最优分割参数。

3.3 地形部位分类

分割之后,需要依据分类规则对分割对象进行判定,进而得出地形部位分类。由于多尺度Geomorphons法分类融入了地形部位多尺度特征,每个栅格单元分类存在由熵值量化的不确定性。为此,本文将每个栅格单元的熵值引入以刻画初始分类中不确定性对分割对象分类的影响,即将熵值作为面积的权重。分割对象分类方法为：将分割对象与初始分类结果叠加,求取待分类对象中每一类地形部位对应指数,依据最大对应指数确定待分类对象所属地形部位类型,即通过权重面积占优,最大对应指数所对应的地形部位类型。当a为分割对象,n为a中栅格单元所属某一地形部位,则n对a的对应指数计算公式为：
K a , n = ∑ i = 1 Num a , n 1 - Entropy a , n , i ) × S ∑ j = 1 M 1 - Entropy a , j ) × S （5）
式中： Entropy ( a , n , i ) 为分割对象a中第i个属于地形部位n的栅格单元对应的熵值; Entropy ( a , j ) 为分割对象a中第j个栅格单元对应的熵值; Num a , n 为分割对象a中属于地形部位n的栅格单元的个数;M为分割对象a中栅格单元的总个数;S为栅格单元面积。

3.4 不确定性计算

在每一分割对象中,通常存在10种初始地形部位中的多种,且每种初始地形部位的占比不同。为较好地评估面向对象地形部位分类结果,熵值可以用来衡量分类结果的不确定性程度：
Ea = - ∑ i = 1 10 K ( a , n ) lo g 2 K ( a , n ) lo g 2 10 （6）
式中：Ea为分割对象a的面向对象分类结果的熵值; K ( a , n ) 为地形部位n对分割对象a中对应指数。显然,Ea的值域范围为[0, 1],其值越大表明分割对象a内地形部位混合度高,面向对象分类不确定性高,反之,其值越小表明分割对象a内地形部位较为单一,面向对象分类不确定性低。因此,通过不确定性分析可反映分类结果的差异性,并具有一定指示作用：不确定性程度高的区域,指示着该区域面向对象分类误差较大且地形多尺度效应较强;不确定性程度低的区域,该区域分类结果的误差一般也相应较低且地形多尺度效应较弱[2]。

4 实验结果与分析

4.1 实验区域与数据

本文选取了陕北黄土高原韭塬沟区域为实验样区,该样区位于无定河中游,隶属黄土峁状丘陵沟壑地貌,丘陵起伏,沟壑纵横,土壤侵蚀极为剧烈。样区海拔814~1188 m,相对高差374 m,地面平均坡度29°,沟壑密度6.52 km/km2。实验数据为国家测绘局生产的1:1万比例尺5 m分辨率DEM和与DEM空间匹配的1 m分辨率DOM数据。样区由4幅邻接的1:1万图幅拼接而成,面积约100 km2。样区位置、样区高程分布以及用于文后实验对比分析的局部区域分布（图3）。
Fig. 3 Location of the study area

图3 研究区位置分布

4.2 结果分析

4.2.1 适宜尺度域分析
在利用均值变点法进行分析时,首先需要确定最大分析尺度（即最大分析窗口）。根据地形部位提取的合理性情况,经过多次实验发现,当分析尺度超过79×79栅格单元时,所得地形部位明显违背地理认知,如山脊部位位于沟沿线以下位置等。至此,由均值变点法得出的平均熵值变化曲线结果如图4所示。从图4可看出,变点位置位于点33处,即在利用多尺度Geomorphons法对样区进行地形部位分类时,适宜分析尺度域为[5×5,33×33]栅格单元。
Fig. 4 The curve of the average entropy

图4 平均熵值变化曲线

基于适宜分析尺度域,地形部位初始分类结果如图5所示。从图5可发现,分类结果与实际地形较为吻合：整体上,山脊部位和沟谷部位交错构成了样区内地形骨架;局部区域,宽阔平坦低洼区被合理分类为平区,山脊与沟谷之间侧向凸起的部位被识别为凸背坡。合理的地形部位分类从侧面印证了均值变点法在确定适宜分析尺度域上的优越性。
Fig. 5 Classification results by Geomorphons method based on multi-scale morphology

图5 多尺度Geomorphons法地形部位分类结果

4.2.2 分割参数分析
本文采用易康（eCongntion）软件中多尺度分割工具对“影像”对象进行多尺度分割。系列实验分析表明分割过程中紧致度和平滑度参数对分割结果影响不显著。因此,本文进行面向对象分割时平滑度紧致度权重各设置为0.5。为了进一步分析分割尺度、光谱权重及形状权重参数组合对分割结果的影响,以0.2为步距,进行了多重参数组合的对比分析实验,结果如图6、表1-2所示。图6表明,当形状权重较小时,分割结果较为破碎,每一个孤立点都分割至一个对象。随着形状权重的增加,孤立点逐渐被分割至“背景”对象中,分割对象也越来越综合。当形状权重增加到一定程度时,本应被分割至不同对象的栅格单元也因综合程度的提升而被分割至同一对象。由表2可知,当形状权重越大,仅由独立栅格单元组成的分割对象占比逐步减小,同时含有多个初始地形部位的分割对象占比逐步增大,即所得分割对象越综合。通过综合对比,当形状权重为0.6时,所得分割结果在有效消除孤立点和含岛对象的同时,也不会因为综合程度过大而影响分割质量,分割对象与实际地表形态特征较为吻合。当形状权重为0.6、紧致度为0.5时,局部方差变率法指示的最优分割尺度为11（图7）。因此,本文基于10波段初始地形部位“影像”,采用分割尺度为11、形状权重为0.6、紧致度权重为0.5的参数组合进行面向对象分割。
Fig. 6 Results of different segmentation parameters （local area No.1）

图6 不同分割参数所得分割结果（局部区域1）

Tab. 1 Different combinations of segmentation parameters

表1 不同的分割参数组合

光谱占比 形状占比 平滑度占比 紧致度占比 最优分割尺度
0.8 0.2 0.5 0.5 11
0.6 0.4 0.5 0.5 10
0.4 0.6 0.5 0.5 11
0.2 0.8 0.5 0.5 17
Tab. 2 Statistic parameters of segmentation objects

表2 分割对象统计数据

独立栅格单元对象占比/% 含多个初始地形部位对象占比/%
形状权重0.2 22.46 64.53
形状权重0.4 17.82 81.65
形状权重0.6 0.00 96.90
形状权重0.8 0.00 98.68
Fig. 7 The curve of the rate of local variance change

图7 局部方差变率曲线

4.2.3 分类结果分析
为了评比权重面积占优与纯面积占优分类规则的优劣,本文基于分割对象,采用上述2种分类规则对地形部位分类结果进行对比分析（图8）。由图8可看出,选中的4个分割对象因分类规则不同而被分类至不同类型地形部位。基于图8（a）,结合图1中各地形部位的定义可以得出,使用权重面积占优分类规则所得分类结果明显比纯面积占优分类规则更加合理。这是由于在分类时纯面积占优规则只考虑待分类对象中各地形部位的面积而忽略了分类的质量（熵值）。综上分析,本文采用权重面积占优分类规则进行分类优势显著。
Fig. 8 Comparison of different classification rules （local area No.2）

图8 不同分类规则结果对比（局部区域2）

图9为采用本文方法和多尺度Geomorphons法所得地形部位分类结果。通过对比可得出,多尺度Geomorphons法的分类结果中地形部位分类较为破碎,其中以地形变化较为剧烈的地区破碎现象尤为明显。为了进一步分析分类结果的完整性,本文使用了平均地形部位斑块面积、地形部位斑块数以及含岛地形部位斑块数来衡量地形部位的完整性,其中地形部位斑块数及含岛地形部位斑块数越大,代表地形部位斑块完整性越差;平均地形部位斑块面积越大,代表着完整性越好。由表3可知,本文面向对象地形部位分类结果的完整性高于多尺度 Geomorphons方法。综上,本文分类结果中地形部位分类破碎问题得到了有效解决,且分类结果更符合地表形态特征以及地理认知。
Fig. 9 Results of two classification methods （local area No.3）

图9 地形部位分类结果（局部区域3）

Tab. 3 Statistic parameters of classification integrity

表3 分类完整性统计

地形部位
斑块数/个		 平均地形部位
斑块面积/m2		 含岛地形部位斑块数/个
≥ 1 ≥2 ≥3 ≥4 ≥5
多尺度Geomorphons法 124 677 779.22 9094 5159 3567 2700 2130
本文方法 25 266 3845.10 1271 561 376 282 225
4.2.4 分类结果不确定性分析
利用熵值计算得出本文方法地形部位分类结果的不确定性程度分布,如图10所示。不确定性空间分布表明,分类结果中熵值较高的地区多位于沟谷等部位,这是因为在沟蚀地貌中,相较于其他地形部位,沟谷地区地形更加复杂多变,从而导致在初始地形部位分类时更容易产生破碎现象,以至面向对象分类结果不确定性偏高。分类结果中熵值较低的地区分布于直背坡和平区,其原因在于,相较于其他地形部位,直背坡和平区地表形态特征更为简单,因而在分类中需要被综合的破碎地形部位也就更少,即相对于而言,这些区域分类不确定性低。然而,值得注意的是,样区内整体熵值偏低,特别是在沟谷与坡面之间的地形复杂区域,该现象表明熵值对于面向对象地形部位分类不确定性指示性较弱,这也与文献[2]所得结论基本一致。
Fig. 10 Uncertainty analysis of terrain position classification

图10 分类不确定性分布

5 结论

本文以地形部分分类为研究视角,提出了多尺度Geomorphons方法适宜分析尺度域的确定方法,构建了顾及多分析尺度的地形部位面向对象分类方法,并以陕北黄土高原区域5 m分辨率DEM为基本数据源,实现了样区内地形部位分类和提取。实验结果表明：① 均值变点法可有效解决分析尺度域难以确定的问题,样区适宜分析尺度域为[5×5, 33×33]栅格单元;② 以0,255为二值化的地形部位数据层组合适用于多尺度分割,尺度、形状及紧致度参数组合影响分割结果,且对于实验样区存在最优分割参数;③ 与多尺度Geomorphons法相比,本文方法得到的地形部位分类结果完整性较好,在地表形态对应和地理认知等方面更具合理性。
然而,不同地形部位的地表形态特征存在一定程度的差异性,其分类需要统筹考虑参数的多样性等方面,特别是适宜分析尺度域与研究区范围、数据源精度等多个因素相关。因此,如何在完善地形部位分类体系的基础上,结合多源地理空间数据（如DOM等）,进一步在分析尺度域精确选取、角度阈值多样性特征、多层次对象分割以及分类精度等方面深入探究,是下一步研究的重点。

The authors have declared that no competing interests exist.

参考文献

原文顺序 | 文献年度倒序 | 文中引用次数倒序
[1]
Hengl T, Reuter H I.Geomorphometry: Concepts, software, applications[M]. Amsterdam: Elsevier, 2008:228-233.

[2]
康鑫,王彦文,秦承志,等.多分析尺度下综合判别的地形元素分类方法[J].地理研究,2016,35(9):1637-1646.地形元素(如山脊、沟谷等)是地表形态类型基本单元,通过地形元素的不同空间组合可形成更高级别的地貌类型.现有的地形元素提取方法大多依靠地形属性计算,难以克服地形元素的空间相关性表达与局部地形属性计算存在不对应的矛盾,Jasiewicz和Stepinski提出的Geomorphons方法——基于高程相对差异信息进行地形元素分类,可避免这一问题,但Geomorphons方法本质上是在单一分析尺度上选择地形特征点用于判别,易受局部地形起伏的影响而造成误分类.针对这一问题,设计出一种多分析尺度下综合判别的地形元素分类方法.应用结果表明:相比Geomorphons方法,利用该方法得到的地形元素的分类结果更为合理.

DOI

[ Kang X, Wang Y W, Qin C Z, et al.A new method of landform element classification based on multi-scale morphology[J]. Geographical Research, 2016,35(9):1637-1646. ]

[3]
Mcdonald R C, Isbell R F, Speight J G, et al.Australian soil and land survey: Field handbook[M]. Melbourne:Inkata Press,1990:17-31.

[4]
江岭,汤国安,赵明伟,等.顾及地貌结构特征的黄土沟头提取及分析[J].地理研究,2013,32(11):2153-2162.沟头是黄土地貌中发育最活跃的地貌部位,沟头的个体与群体对黄土沟间地的蚕食,成为监测黄土高原地面侵蚀演化的重要标识.有效获取并分析沟头的空间分布特征,对于系统、深入地研究黄土地貌的空间形态结构具有重要的意义.以陕北黄土高原典型地貌类型区为实验样区,设计并实现了基于5m分辨率DEM的沟头提取算法,获取了各实验样区的沟头空间信息;运用点格局分析方法,分析了沟头的空间分布特征.实验结果显示,所提出顾及沟沿线的沟头自动提取方法,具有较好的合理性与精确性;同时,从黄土塬区、黄土残塬区到黄土丘陵沟壑区,随着沟壑发育程度的增强,沟头逐步逼近分水线,其空间分布呈现“远分水线聚集—随机—近分水线聚集”的变化格局.

DOI

[ Jiang L, Tang G A, Zhao M W, et al.Extraction and analysis of loess gully heads considering geomorphological structures[J]. Geographical Research, 2013,32(11):2153-2162. ]

[5]
秦承志,卢岩君,邱维理,等.模糊坡位信息在精细土壤属性空间推测中的应用[J].地理研究,2010,29(9):1706-1714.

[ Qin C Z, Lu Y J, Qiu W L, et al.Application of fuzzy slope positions in predicting spatial distribution of soil property at finer scale[J]. Geographical Research, 2010,29(9):1706-1714. ]

[6]
MacMillan R A, Moon D E, Coupé R A. Automated predictive ecological mapping in a forest region of B C, Canada, 2001-2005[J]. Geoderma, 2007,140(4):353-373.Procedures for automated predictive thematic mapping were developed and applied to project areas totaling more than 3million ha of forested land in British Columbia, Canada. The effective scale of mapping was 1:20,000 using data at a grid resolution of 25m. The methods can be described as a form of automated feature extraction or object recognition where the objects of interest consist of ecological site types. The methods implement a hybrid of automated, semi-automated and manual procedures that develop and apply heuristic, rule-based conceptual models of ecological-landform relationships. The methods rely heavily upon terrain derivatives extracted from available digital elevation models (DEMs) in addition to satellite imagery and manually digitized maps of ancillary environmental conditions. The primary input has been the BC provincial Terrain Resource Information Management (TRIM) digital elevation model (DEM) surfaced to a regular grid of 25m. Other input layers include manually interpreted maps of parent material texture, depth and ecological exception classes, manually prepared maps of the spatial distribution of ecological zones of the BC Biogeoclimatic Ecosystem Classification (BEC) system and, to a limited extent, LandSat7 digital satellite imagery. The procedures do not use any field sampling to develop or train classification rules. A knowledge-based approach is used to establish classification rules which are defined and implemented using a Semantic Import (SI) Model implementation of fuzzy logic. All rules are constructed by examining and deconstructing published field guides that define the required ecological output classes and that document the current expert understanding of the conditions and criteria that control the spatial distribution of these desired output classes. An iterative, trial and error, process is used to develop, apply, evaluate and revise object recognition rules that relate ranges of values of key input data layers to an expert-assigned likelihood of occurrence for each ecological class of interest. Local expert knowledge is used at each stage to evaluate each new set of output results and to guide refinement of the fuzzy SI model classification rules. Field sample observations obtained along randomly selected closed traverses were collected following a line intercept approach and used to assess the accuracy of the final predictions of ecological classification. Application of the procedures has progressed from an initial pilot project through projects to evaluate operational scale-up to full-scale commercial application to millions of hectares. Costs have been reduced from a high of $3.50/ha to less than $0.20/ha. Rates of progress increased from 150,000ha per person year to more than 2.0million ha per person year. Independent assessments of map accuracy produced results superior to the highest accuracies reported for all alternatives, including traditional manual mapping methods. We conclude that we have formalized and automated many of the concepts and techniques previously used to create thematic maps of ecosystems using manual interpretation of stereo air photos and ancillary data combined with field observations. We have shown that automated feature extraction is able to capture and apply the concepts of landform control referenced by typical landform-based ecological models and classification systems. We have demonstrated that it is possible to produce accurate and cost-effective ecological-landform maps by applying fuzzy and Boolean logic and automated landform analysis procedures to widely available spatial data.

DOI

[7]
李文杰,王兴奎,李丹勋,等.基于物理过程的分布式流域水沙预报模型[J].水利学报,2012,43(3):264-274.将分布式水文模型与土壤侵蚀产沙模型相互嵌套,构建了一个分布式流域水沙预报模型。模型基于DEM,将流域离散为＂坡面＋沟道＂系统,采用＂＋1＂分级方法将沟道组织为河网,实现了流域水沙预报的并行计算。坡面单元划分为植被层、地表层、非饱和带和地下水层,模拟了降雨、截留、蒸散发、入渗、地表径流、壤中流和地下水、各种形式的侵蚀产沙及泥沙输移;沟道进行汇流演算以及泥沙输移。模型在李子溪流域对长系列降雨的模拟结果表明,模型的水文模块模拟精度良好,侵蚀产沙模块模拟效果基本令人满意。

[ Li W J, Wang X K, Li D X, et al.A physically-based distributed model for the prediction of watershed water erosion[J]. Journal of Hydraulic Engineering, 2012,43(3):264-274. ]

[8]
王彦文,秦承志.地貌形态类型的自动分类方法综述[J].地理与地理信息科学,2017,33(4):16-21.

[ Wang Y W, Qin C Z.A review on methods for landform automatic classification[J]. Geography and Geo-information Science, 2017,33(4):16-21. ]

[9]
Wieczorek M, Migoń P.Automatic relief classification versus expert and field based landform classification for the medium-altitude mountain range, the sudetes, S W Poland[J]. Geomorphology, 2014,206:133-146.The main purpose of the study is to compare the results of automatic classification of landforms performed as unsupervised classification and manual (expert) morphographic landform classification derived from field mapping and recognition of long-term trends in landform evolution. Two test areas in the Sudetes Mountains (SW Poland) were selected, each representing a different type of relief. The Jelenia Góra Basin is an example of denudational hilly granite topography, with inselbergs, other residual hills, and basins, whereas the northern part of the Bystrzyckie Mountains is a faulted plateau underlain by sedimentary and metamorphic rocks. The setup of the classification involved a DEM derived from digitalization of 1:25,000 maps, a filter mask of 2002×022002pixels (50002×0250002m), k -median cluster analysis, and consideration of four dimensions (relative height, slope, combined curvature, and aspect). We used a statistical criterion to decide the level of classification, i.e. the number of homogeneous groups returned. Output maps show the results of classification into five (Jelenia Góra Basin) or six groups (Bystrzyckie Mountains), later confronted with nine-group classification. In both areas automatic classification for k 02=02502÷026 performed well and the principal features of relief are well captured, despite very different landform patterns and long-term geomorphic histories. Slope and curvature are the main variables to differentiate individual terrain units (landforms), with an additional role of aspect in the faulted and tilted plateau. For k 02>026 aspect becomes the main variable in the hilly relief and the pattern of terrain units becomes more fragmented, adversely impacting on the clarity of interpretation, while surfaces in the faulted plateau setting are further subdivided according to slope and curvature. The match between ‘expert’ maps and the results of automatic classification varies, depending on the landforms considered. While certain landforms are clearly associated with one particular group yielded by automatic classification, others fall into a number of groups and do not have distinctive signatures, despite different origins.

DOI

[10]
Toshiya M, Masamu A. Automated segmentation of hillslope profiles across ridges and valleys using a digital elevation model[J]. Geomorphology, 2012,177-178:167-177.78 We developed an algorithm to automate hillslope-profile segmentation using a DEM. 78 The algorithm was examined in fluvial landscapes of Japan. 78 The segmentation results agreed well with the manual interpretation of contours.

DOI

[11]
Drăguţ L, Minar J. Land-surface segmentation to delineate elementary forms from digital elevation model[R]. In: The Proceedings of Geomorphometry 2013, Nanjing, 2013:61-64.

[12]
Minár J, Evans I S.Elementary forms for land surface segmentation: The theoretical basis of terrain analysis and geomorphological mapping[J]. Geomorphology, 2008,95(3):236-259.The wealth of types of elementary forms and their boundaries is potentially unbounded and thus is sufficient to cover the real variety of landforms. Elementary forms in the basic set proposed here have clear potential for genetic and dynamic interpretation. A brief worked example documents the possibility of analytical computation of various models of ideal elementary forms for particular segments of landform. Ideal elementary forms can be considered as attractors, to which the affinity of surface segments can be measured by multivariate statistical methods. The use of the concept of elementary forms in landscape segmentation is promising and it could be adapted for elementary segmentation of various other spatial fields.

DOI

[13]
秦承志,朱阿兴,施迅,等.坡位渐变信息的模糊推理[J].地理研究,2007,26(6):1165-1175.

[ Qin C Z, Zhu A X, Shi X, et al.Fuzzy inference of spatial gradation of slope position[J]. Geographical Research, 2007,26(6):1165-1175. ]

[14]
秦承志,朱阿兴,李宝林,等.坡位的分类及其空间分布信息的定量化[J].武汉大学学报·信息科学版,2009,34(3):374-377.

[ Qin C Z, Zhu A X, Li B L, et al.Taxonomy of slope positions and quantification of their spatial distribution information[J]. Geomatics and Information Science of Wuhan University, 2009,34(3):374-377. ]

[15]
Qin C Z, Zhu A X, Shi X, et al.Quantification of spatial gradation of slope positions[J]. Geomorphology, 2009,110(3):152-161.Transition between slope positions (e.g., ridge, shoulder slope, back slope, foot slope, and valley) is often gradual. Quantification of spatial transitions or spatial gradations between slope positions can increase the accuracy of terrain parameterization for geographical or ecological modeling, especially for digital soil mapping at a fine scale. Current models for characterizing the spatial gradation of slope positions based on a gridded DEM either focus solely on the parameter space or depend on too many rules defined by topographic attributes, which makes such approaches impractical. The typical locations of a slope position contain the characteristics of the slope position in both parameter space and spatial context. Thus, the spatial gradation of slope positions can be quantified by comparing terrain characteristics (spatial and parametrical) of given locations to those at typical locations. Based on this idea, this paper proposes an approach to quantifying the spatial gradation of slope positions by using typical locations as prototypes. This approach includes two parts: the first is to extract the typical locations of each slope position and treat them as the prototypes of this position; and the second is to compute the similarity between a given location and the prototypes based on both local topographic attributes and spatial context. The new approach characterizes slope position gradation in both the attribute domain (i.e., parameter space) and the spatial domain (i.e., geographic space) in an easy and practicable way. Applications show that the new approach can quantitatively describe spatial gradations among a set of slope positions. Comparison of spatial gradation of A-horizon sand percentages with the quantified spatial gradation of slope positions indicates that the latter reflects slope processes, confirming the effectiveness of the approach. The comparison of a soil subgroup map of the study area with the maximum similarity map derived from the approach also suggests that the quantified spatial gradation of slope position can be used to aid geographical modeling such as digital soil mapping.

DOI

[16]
Pennock D J, Zebarth B J, De Jong E.Landform classification and soil distribution in hummocky terrain, Saskatchewan, Canada[J]. Geoderma, 1987,40(3):297-315.

DOI

[17]
Jasiewicz J, Stepinski T F.Geomorphons: A pattern recognition approach to classification and mapping of landforms[J]. Geomor Phology, 2013,182:147-156.We introduce a novel method for classification and mapping of landform elements from a DEM based on the principle of pattern recognition rather than differential geometry. At the core of the method is the concept of geomorphon (geomorphologic phonotypes) — a simple ternary pattern that serves as an archetype of a particular terrain morphology. A finite number of 498 geomorphons constitute a comprehensive and exhaustive set of all possible morphological terrain types including standard elements of landscape, as well as unfamiliar forms rarely found in natural terrestrial surfaces. A single scan of a DEM assigns an appropriate geomorphon to every cell in the raster using a procedure that self-adapts to identify the most suitable spatial scale at each location. As a result, the method classifies landform elements at a range of different spatial scales with unprecedented computational efficiency. A general purpose geomorphometric map — an interpreted map of topography — is obtained by generalizing allgeomorphons to a small number of the most common landform elements. Due to the robustness and high computational efficiency of the method high resolution geomorphometric maps having continental and even global extents can be generated from giga-cell DEMs. Such maps are a valuable new resource for both manual and automated geomorphometric analyses. In order to demonstrate a practical application of this new method, a 30m cell611 geomorphometric map of the entire country of Poland is generated and the features and potential usage of this map are briefly discussed. The computer implementation of the method is outlined. The code is available in the public domain.

DOI

[18]
Andrew D. topographic position and landforms analysis, the nature conservancy[OL]. , 2001.

[19]
马振华,李小苗, 本泓,等.青藏高原东北缘马衔山夷平面特征指数的提取与分析[J].地理学报,2016,71(3):400-411.

[ Ma Z H, Li X M, Guo B H, et al.Extraction and analysis of Maxianshan planation surfaces in northeastern margin of the Tibetan Plateau[J]. Acta Geographica Sinica, 2016,71(3):400-411. ]

[20]
范建容,张子瑜,李立华.四川省山地类型界定与山区类型划分[J].地理研究,2015,34(1):65-73.山地类型的界定是科学认知山地规律的基础,对山地科学的发展有着基础性的意义。山地科学研究的出发点与落脚点,是为了促进山区区域在各方面的可持续发展,所以在山地类型界定的研究结果基础上对山区类型进行划分,是山区可持续发展的基本科学依据。采用GIS技术并结合DEM数据,对四川省的山地类型进行界定,同时对山区的类型进行划分。在山地类型界定涉及到的地形起伏度计算中,对均值变点法的应用结果进行了讨论与改进,得到地形起伏度最佳统计单元为9.92 km2。将四川省山地类型界定为丘陵、低山、中低山、中山、次高山、高山、极高山;将山区划分为纯丘陵县、半山区县、准山区县、显山区县、整山区县,并对山区发展的政策方向提出了一定的建议。

DOI

[ Fan J R, Zhang Z Y, Li L H.Mountain demarcation and mountainous area divisions of Sichuan Province[J]. Geographical Research, 2015,34(1):65-73. ]

[21]
周启鸣,刘学军,数字地形分析[M].北京:科学出版社,2006:45.

[ Zhou Q M, Liu X J.Digital terrain analysis[M]. Beijing: Science Press, 2006:45. ]

[22]
常直杨,孙伟红,王建,等.青藏高原及其邻近地区地貌类型划分[J].山地学报,2017,35(1):1-8.我国至今尚未形成一个公认的地貌分类系统,传统地貌类型的划分主要依据海拔及起伏度,较少考虑地貌的完整性原则,且分类结果琐碎。为了避免此类问题,以我国青藏高原及其邻近地区1000 m分辨率的SRTM DEM为数据源,采用面向对象思想,基于eCognition软件,利用多尺度分割、局部方差法以及决策树分类法自动划分了地貌形态。结果表明：（1）在分割尺度范围为10-1400、步长为100时,青藏高原及其邻近地区的最佳分割尺度为400;（2）依据平均高程及标准差的大小,青藏高原及其邻近地区可划分为极高山、大起伏高山、高丘、小起伏高山、大起伏中山、小起伏中山、高海拔平地、低海拔平地八种地貌类型。相比依据海拔及起伏度的划分方法,分类结果更能考虑地貌的完整性原则,且具有高效便捷性,划分结果更平滑,为我国地貌类型的划分提供了参考。

DOI

[ Chang Z Y, Sun W H, Wang J, et al.Object-oriented method based on classification of geomorphic type in the Tibet Plateau and adjacent regions[J]. Mountain Research, 2017,35(1):1-8. ]

[23]
Lucian D, Dirk T, Shaun R L.ESP: A tool to estimate scale parameter for multiresolution image segmentation of remotely sensed data[J]. International Journal of Geographical Information Science, 2010,24(6):859-871.The spatial resolution of imaging sensors has increased dramatically in recent years, and so too have the challenges associated with extracting meaningful information from their data products. Object-based image analysis (OBIA) is gaining rapid popularity in remote sensing science as a means of bridging very high spatial resolution (VHSR) imagery and GIS. Multiscalar image segmentation is a fundamental step in OBIA, yet there is currently no tool available to objectively guide the selection of appropriate scales for segmentation. We present a technique for estimating the scale parameter in image segmentation of remotely sensed data with Definiens Developer. The degree of heterogeneity within an image-object is controlled by a subjective measure called the cale parameter , as implemented in the mentioned software. We propose a tool, called estimation of scale parameter (ESP), that builds on the idea of local variance (LV) of object heterogeneity within a scene. The ESP tool iteratively generates image-objects at multiple scale levels in a bottom-up approach and calculates the LV for each scale. Variation in heterogeneity is explored by evaluating LV plotted against the corresponding scale. The thresholds in rates of change of LV (ROC-LV) indicate the scale levels at which the image can be segmented in the most appropriate manner, relative to the data properties at the scene level. Our tests on different types of imagery indicated fast processing times and accurate results. The simple yet robust ESP tool enables fast and objective parametrization when performing image segmentation and holds great potential for OBIA applications.

DOI

Options
文章导航

/

〈 〉