EN中文

Journal of Geo-information Science >

2016 , Vol. 18 >Issue 5: 632 - 638

DOI: https://doi.org/10.3724/SP.J.1047.2016.00632

Orginal Article

Optimal Scales Based Segmentation of High Spatial Resolution Remote Sensing Data

  • YANG Haiping , 1, 2, * ,
  • MING Dongping 3
Expand
  • 1. College of Computer Science & Technology, Zhejiang University of Technology, Hangzhou 310023, China
  • 2. Key Laboratory of Oceanographic Big Data Mining & Application of Zhejiang Province, Zhoushan 316022, China
  • 3. School of Information Engineering, China University of Geosciences (Beijing), Beijing 100083, China;
*Corresponding author: YANG Haiping, E-mail:

Received date: 2016-02-15

  Request revised date: 2016-04-18

  Online published: 2016-05-10

Copyright

《地球信息科学学报》编辑部 所有
Fold

Abstract

The quality of image segmentation has a great impact on the results of information extraction from high spatial resolution remote sensing imagery when the object-based method is employed. During the segmentation of high spatial resolution remote sensing images, the scale parameter directly affects the construction of segmented image objects. A small scale is likely to produce broken image objects, while a large scale probably results in the mixed image objects. To solve this problem, an image segmentation framework based on a set of optimal scales is proposed in this paper. First of all, the high spatial resolution remote sensing image is processed using multi-scale segmentation methods with respect to a group of regularly distributed scales. Then the relationship between the global standard deviation of a single segmented layer and its corresponding scale is determined, from which a group of optimal scales are selected. Since the object in a layer that is segmented by a big scale parameter contains the corresponding object in a layer that is segmented by a small scale parameter, a hierarchical tree with nodes of multi-scale image objects can be created. Within this hierarchical tree, the image object of the layer that is segmented by the maximum scale is set as the root. In this manner, each image object of the layer that is segmented by the maximum scale can generate a hierarchical tree, which all together forms the image forest. Two types of features are considered when the optimal image object is selected from each hierarchical tree, which are the comprehensive evaluation index and the spectral features. The comprehensive evaluation index keeps a balance between the homogeneity and heterogeneity of the image objects. And the spectral features of the children nodes should be consistent with the parent nodes in order to dismiss the mixed image objects. Finally, the segmented result is created after the optimal image objects from all hierarchical trees are selected. In the experiment presented in this paper, the Geoeye and ZY3 images are adopted. Results show that the proposed method can effectively improve the percentage of properly segmented image objects.

Key words： high spatial resolution remote sensing imagery; segmentation; multiple scales; a hierarchical tree with nodes of multi-scale image objects; comprehensive evaluation index

Cite this article

YANG Haiping , MING Dongping . Optimal Scales Based Segmentation of High Spatial Resolution Remote Sensing Data[J]. Journal of Geo-information Science, 2016 , 18(5) : 632 -638 . DOI: 10.3724/SP.J.1047.2016.00632

1 引言

从20世纪末开始至今,能够获取米级甚至亚米级空间分辨率影像的卫星不断发射升空,如国外的Geoeye系列、Worldview系列、Pleiades系列,以及国内的高分（GF）系列、资源（ZY）卫星系列等。这些高分辨率遥感影像在减灾、国土、生态等领域发挥越来越重要的作用。因此,如何有效地从高分辨率影像中提取行业应用所需的信息是高分辨率影像能否在应用中发挥作用的关键问题。针对高分辨率影像而言,与传统的中低分辨率影像采用的基于像元的处理方法相比,面向对象的处理方法能够有效地避免处理过程中出现的椒盐噪声等问题[1-4]。而影像分割是面向对象处理方法中最基本的步骤。
近十几年来,随着影像空间分辨率的提高,遥感影像分割算法有了快速的发展,如神经网络模型、分水岭算法、均值漂移分割算法等都被应用到遥感影像的分割中[5-10]。已有的研究表明,涉及到实际分割操作实验时,被集成到商业软件eCognition中的Multi-resolution模型出现的频率较高,它作为实验的一个分割步骤或在对比实验中出现[11-14]。这些分割算法中,涉及到的一个重要问题是尺度选择。不同的地物（如耕地、湖泊、农村房屋、城市公寓等）需要不同的分割尺度,无法采用一个统一的尺度进行描述。如果采用小尺度分割,那么势必产生过分割现象;如果采用一个大尺度分割,那么可能产生欠分割的问题,从而出现混合对象,如道路和行道树组成一个对象,房屋和道路组成一个对象等。可见,分割尺度的不同将对分割结果产生很大的影响。对此,研究人员提出了多尺度分割方法,包括“自底向上”和“自顶向下”2种策略。在不同尺度的合并过程中,都需要有一个阈值来判断是否进行合并。例如,eCognition软件中的Multi-resolution分割算法采用了“自底向上”合并的策略,用户可以通过控制尺度参数来调整合并阈值,尺度越大,分割得到的影像对象就越大[15]。
从目前已有的多尺度分割方法来看,大多数方法在每次分割时主要采用一种尺度。本文针对高分辨率影像分割中地物多尺度的问题,将在一次分割中同时综合多个尺度的分割结果。实验结果表明,本文的算法能够有效地提高正常分割影像对象的比例。

2 基于多层优选尺度的分割方法

首先给出多层次对象树和影像森林的定义。多层次对象树是由同一地物的不同尺度分割对象构成的树,尺度最大的分割对象为多层次对象树的根节点,尺度最小的分割对象为多层次对象树的叶节点。由所有多层次对象树组成的图定义为影像森林。
基于多层优选尺度的高分辨率影像分割流程如图1所示。首先,对高分辨率影像进行多尺度分割,同时计算分割对象的特征。在此基础上,分别计算各尺度的影像整体特征,通过影像特征变化率与尺度的关系选择一系列最优分割结果。以尺度最大的优选分割对象为根节点,自上而下,形成一系列多层次对象树,由此来构建影像森林。然后,计算影像对象的局部综合评价参数。最后,分别遍历影像森林中的多层次对象树,逐个输出局部最佳影像对象,合并成为由多个尺度综合形成的分割结果。
Fig.1 Process of image segmentation based on a series of optimal scales

图1 基于多层优选尺度的高分影像分割方法

2.1 最优分割尺度的选择

按照最优尺度选择的数目,最优尺度选择方法可以分为单尺度和多尺度2种。单尺度的方法是在一景影像中选择一个最优尺度[16],在实际应用中,这种单尺度选择方法无法满足各类地物的不同尺度要求。多尺度选择的方法中,按照研究对象的不同,可以分为针对地物类别的最优尺度选择和针对影像的最优尺度选择。按照不同地物类别选择尺度的方法可参见文献[17]、[18],这类方法适合用于专门提取某一类地物信息的情况。影像选择最优尺度的方法包括将影像分区选择最优尺度[19]以及直接在整景影像上选择最优尺度[20-21]2种类型。通过影像分区选择不同区域的最优尺度时需要充分考虑分区的大小及分区内的地物类别,否则会产生单尺度选择法中遇到的问题,而对于影像全局范围内直接选择多个分割尺度的方法则无需考虑这些问题。因此,本文将采用在全局范围内综合选择多个尺度的策略。
从影像全局而言,当混合对象增多时,相邻对象之间的光谱异质降低,影像中所有对象的均值方差变小;反之,当混合对象减少时,则相邻对象之间的光谱差异增大,所有对象的均值方差增大[20]。研究表明,最优分割的参考值发生在均值方差峰值且方差变化率开始呈现下降趋势的分割值处[20-21]。本文采用了影像分割层全局标准差的变化与尺度的关系来确定一组最优分割尺度[20-21],如式（1）所示。
Δ = WS D L - WS D L - 1 WS D L - 1 , L > 0 （1）
式中： L 表示分割层; WSD 表示影像全局的加权标准差,公式如式（2）所示。
WSD = ∑ i = 1 m w i S D i w i = are a i A S D i = MAX ( S D ij ) , j = 1,2 , ... , n （2）
式中： S D i 表示第i个影像对象的波段标准差; w i 表示第i个影像对象的面积权重;m表示在当前分割层的影像对象总数;n表示影像波段数目。从加权标准差变化率-尺度关系图中可以选出多个最优尺度,即局部对应的峰值为相应的最优分割尺度。

2.2 多层次对象树的构建

本文以三层对象树为例对多层次对象树的构建进行说明。假设现有影像I通过自底向上的分割方法获得了一组从小到大排列为S1、S2、S3的分割结果（图2）。尺度S1、S2、S3中相应位置的影像对象 O S 1 、 O S 2 、 O S 3 满足式（3）的父子包含关系。
O S 1 ⊆ O S 2 O S 2 ⊆ O S 3 （3）
Fig.2 Diagram of a hierarchical tree with nodes of multiscale image objects

图2 多层次对象树示意图

本文以尺度最大( S 3 )的分割层中的对象 { O i | O i ∈ O S 3 , i = 1,2 , ⋯ , N S 3 } 为根节点, N S 3 为 S 3 分割层中影像对象的数目,寻找尺度S2的分割层中和对象 { O i | O i ∈ O S 3 , i = 1,2 , ⋯ , N S 3 } 具有相等或被包含关系的对象 { O i | O i ∈ O S 2 , i = 1,2 , ⋯ , N S 2 } ,选中该对象并将其作为对象 { O i | O i ∈ O S 3 , i = 1,2 , ⋯ , N S 3 } 的子节点。依次类推,在尺度S1的分割层中寻找和对象 { O i | O i ∈ O S 2 , i = 1,2 , ⋯ , N S 2 } 具有相等或被包含关系的对象 { O i | O i ∈ O S 1 , i = 1,2 , ⋯ , N S 1 } ,并作为对象 { O i | O i ∈ O S 2 , i = 1 , 2 , ⋯ , N S 2 } 的子节点。通过以上方法构建的树就是多层次对象树。以S3中的分割对象为根节点,分别构建相应的多层次对象树。所有这些多层次对象树形成了影像森林。

2.3 多层次对象树的合并

本文从局部对多层次对象树进行合并,且合并时综合考虑了图像本身的特性及地物的光谱特征。图像本身的特性包括影像对象的同质性及异质性,理想状况下,希望分割对象的同质性越大同时异质性越小。本文提出了一个融合了同质性及异质性的综合评价指数 ε ,如式（4）所示。
ε = h m 2 + 1 / h t 2 + 1 2 （4）
式中： h m 表示同质性的归一化量度; h t 表示异质性的归一化量度,具体计算如式（5）、（6）所示。
h m = MAX ( h mo ) - h mo MAX ( h mo ) - MIN ( h mo ) + 1 （5）
h t = MAX ( h te ) - h te MAX ( h te ) - MIN ( h te ) + 1 （6）
式中： MAX ( ∙ ) 表示取最大值函数; MIN ( ∙ ) 取最小值函数; h mo 表示同质性量度; h te 表示异质性量度。本文采用像元标准差作为同质性量度,其值越小说明该对象内部像元波动越小,即同质性越好;光谱均值之差作为异质性量度,其值越大说明对象之间异质性越大。结合式（4）,当影像对象的同质性越大（即标准差越小）、异质性越大时, ε 越小,因此本文选择 ε 最小的对象作为候选最佳对象。
在算法自底向上搜索的过程中,以最底层优选尺度中分割对象的光谱特征为基准,给上层最优尺度中分割对象的光谱特征加上限制条件,即在通过综合评价指数 ε 选择了最佳对象之后,要求上层分割对象的地物光谱特征必须和最底层对应分割对象的光谱特征保持一致,从而避免出现混合对象。本文中地物光谱特征主要考虑了植被和水体2类特征明显的地物。

3 实验与分析

3.1 Geoeye多光谱影像分割实验

本实验中采用了Geoeye多光谱影像,实验数据的空间分辨率为2 m,影像大小1000像元×1000像元,有蓝（450~510 nm）、绿（510~580 nm）、红（655~690 nm）和近红外（780~920 nm）共4个波段。该研究区域（图3）的地物类型以耕地、水体、林地为主体,同时还分布有一些村庄。
Fig.3 Geoeye multispectral image

图3 Geoeye多光谱影像

首先,采用eCognition Developer 8.7的多分辨率算法对Geoeye影像进行分割,除了尺度以外的参数都按照算法的默认设置,分割尺度范围设为5-120,间隔设为1。根据2.1节中的方法,由标准差变化率-尺度曲线得到一组Geoeye影像的最优分割尺度,即24、35、49、59、69、94、116。据此,实验将以尺度116的分割对象为根节点,构建一系列深度为7的多层次对象树以形成影像森林。然后,比较多层次对象树中每个节点的综合评价指数 ε ;同时,通过采用NDVI及NDWI限制父层对象的光谱特性,来搜寻局部最优的影像对象。
本文通过过分割、欠分割及正常分割对象的比例来评价分割质量,随机抽取N个分割对象,计算过分割对象的比例（POS）、欠分割对象的比例（PUS）及正常分割对象的比例（PS）,如式（7）-（9）所示。
P OS = N OS N （7）
P US = N US N （8）
P S = N S N （9）
式中：NOS表示抽样中过分割影像对象的总数;NUS表示抽样中欠分割影像对象的总数;NS表示抽样中正常分割影像对象的总数。
从Geoeye多光谱影像分割结果中随机抽取200个影像对象评价分割算法的质量,分别计算本算法及各优选尺度分割结果的过分割、欠分割及正常分割影像对象的比例,结果如图4所示。图4中横轴表示不同的分割方法,其中C表示本文算法的评价结果,116-24表示相应尺度分割算法的评价结果。从实验结果可知,本文算法获得的正常分割结果比例最高,为56.50%,其欠分割影像对象占27.50%,过分割影像对象占16.00%。在7个优选尺度中,尺度49的正常分割结果最高,为40.00%,尺度116的正常分割结果最低,为16.00%;尺度116的分割结果中欠分割对象比例最大,为82.00%,尺度24的分割结果中欠分割对象的比例最小,为3.50%;尺度116的分割结果中过分割对象的比例最大,为63.00%;尺度24的分割结果中过分割对象的比例最低,为2.00%。
Fig.4 Comparison of the segmentation results using different methods for the Geoeye multispectral image

图4 Geoeye多光谱影像分割实验中不同尺度分割结果和本文方法分割结果的精度比较

图5所列为Geoeye多光谱影像分割实验中Geoeye多光谱影像局部的分割结果,图5（a）表示本文算法的分割结果,图5（b）-（h）分别表示尺度116-24的分割结果。从目视的角度而言,图5（a）不仅将建筑用地作为一个整体分割出来,而且能将林地和耕地分割开来,而图5（b）（尺度116）、图5（c）（尺度94）、图5（d）（尺度69）的分割结果中建筑用地、林地及耕地都合成了一个大对象,图5（e）（尺度59）、图5（f）（尺度49）、图5（g）（尺度35）的分割结果中林地及耕地合成了一个对象,图5（h）（尺度24）分割结果比较破碎。
Fig.5 Details of the segmentation results for the Geoeye multispectral image

图5 Geoeye多光谱影像分割实验的局部结果图

3.2 ZY3多光谱影像分割实验

本实验采用了ZY3多光谱影像进行分割,实验数据的空间分辨率为5.8 m,影像大小为3000像元×3000像元,有蓝（450~520 nm）、绿（520~590 nm）、红（630~690 nm）和近红外（770~890 nm）共4个波段。该研究区域（图6）的地物类别包括耕地、水体、林地,还有一大一小2个城镇分布。首先,将ZY3影像输入eCognition Developer 8.7中的多分辨率分割算法,除了尺度以外的参数都按照算法的默认设置,分割尺度范围设为10-65,间隔设为1。然后计算标准差变化率-尺度曲线,并从中得到该ZY3多光谱影像的一组最优分割尺度为19、28、31、34、41、44、51、62。由此,实验将以尺度62的分割对象为根节点,构建一系列深度为8的多层次对象树以形成影像森林。其余处理方法与上述3.1节中的实验相同。
Fig.6 ZY3 multispectral image

图6 ZY3影像

从ZY3影像的分割结果中随机选择了200个影像对象来评价分割结果的质量,分别计算本文算法及各优选尺度分割结果的过分割、欠分割及正常分割影像对象的比例,结果如图7所示。图7中横轴表示不同的分割方法,其中C表示本文算法的评价结果,62-19表示相应的分割尺度所对应的评价结果。从实验结果可知,由本文算法获得的正常分割结果比例最高,为59.50%,其欠分割影像对象占27.00%,过分割影像对象占13.50%。在8个优选尺度中,尺度19的正常分割结果最高,为31.00%,尺度51的正常分割结果最低,为20.00%;尺度19的分割结果中欠分割对象的比例最低,为7.00%,尺度62的分割结果中欠分割对象的比例最高,为64.00%;尺度19的分割结果中过分割对象的比例最高,为62.00%,尺度62的分割结果中过分割对象的比例最低,为10.50%。
Fig.7 Comparison of the segmentation results using different methods for the ZY3 multispectral image

图7 ZY3多光谱影像分割实验中不同尺度分割结果和本文方法分割结果的精度比较

图8所列为ZY3多光谱影像分割实验中厂房及其周围道路局部的分割结果,图8（a）表示本文算法的分割结果,图8（b）-（i）分别表示尺度62-19的分割结果。从目视的角度而言,图8（a）不仅将厂房作为一个整体分割出来,而且能将厂房前的绿化带分割开来,而图8（b）（尺度62）、图8（c）（尺度51）、图8（d）（尺度44）的分割结果中厂房、绿化带及道路都合成了一个混合对象,图8（e）（尺度41）、图8（f）（尺度34）、图8（g）（尺度31）、图8（h）（尺度28）、图8（i）（尺度19）的分割结果中单个厂房被分割为多个对象。但由于在最小的尺度图8（i）（尺度19）中,小厂房和周围的道路被分割为一个混合对象,所以本文算法结果中也出现了该混合对象。
Fig.8 Details of the segmentation results of factory buildings for the ZY3 multispectral image

图8 ZY3多光谱影像分割实验的局部结果图

4 结语

本文针对高分辨率影像分割中地物多尺度的问题,提出了一种基于多层优选尺度的高分辨率影像分割算法。由Geoeye多光谱影像分割实验和ZY3多光谱影像分割实验的结果可知,本文算法得到的分割结果中正常影像对象的比例高于任一单一优选分割尺度的分割结果。本文算法通过构建多层次对象树,将单尺度分割结果作为树的一层节点参与计算,并采用了一个融合了同质性及异质性的综合评价指数 ε 在局部选择最优的分割影像对象。与大的单尺度分割相比,其能够避免大量的欠分割影像对象;与小的单尺度分割相比,其能够避免大量的过分割影像对象。
从算法效率角度而言,虽然算法中已经采用了空间索引等方法进行加速,但是本文算法的效率还有待进一步提高。在今后的工作中,将引入并行计算方法以进一步提高算法效率。

The authors have declared that no competing interests exist.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Li X, Shao G.Object-based land-cover mapping with high resolution aerial photography at a county scale in midwestern USA[J]. Remote Sensing, 2014,6(11):11372-11390.

[2]
Myint S W, Gober P, Brazel A, et al.Per-pixel vs. object-based classification of urban land cover extraction using high spatial resolution imagery[J]. Remote Sensing of Environment, 2011,115(5):1145-1161.In using traditional digital classification algorithms, a researcher typically encounters serious issues in identifying urban land cover classes employing high resolution data. A normal approach is to use spectral information alone and ignore spatial information and a group of pixels that need to be considered together as an object We used QuickBird image data over a central region in the city of Phoenix, Arizona to examine if an object-based classifier can accurately identify urban classes. To demonstrate if spectral information alone is practical in urban classification, we used spectra of the selected classes from randomly selected points to examine if they can be effectively discriminated. The overall accuracy based on spectral information alone reached only about 63.33%. We employed five different classification procedures with the object-based paradigm that separates spatially and spectrally similar pixels at different scales. The classifiers to assign land covers to segmented objects used in the study include membership functions and the nearest neighbor classifier. The object-based classifier achieved a high overall accuracy (90.40%), whereas the most commonly used decision rule, namely maximum likelihood classifier, produced a lower overall accuracy (67.60%). This study demonstrates that the object-based classifier is a significantly better approach than the classical per-pixel classifiers. Further, this study reviews application of different parameters for segmentation and classification, combined use of composite and original bands, selection of different scale levels, and choice of classifiers. Strengths and weaknesses of the object-based prototype are presented and we provide suggestions to avoid or minimize uncertainties and limitations associated with the approach. (C) 2011 Elsevier Inc. All rights reserved.

DOI

[3]
Chen Y, Su W, Li J, et al.Hierarchical object oriented classification using very high resolution imagery and LIDAR data over urban areas[J]. Advances in Space Research, 2009,43(7):1101-1110.<h2 class="secHeading" id="section_abstract">Abstract</h2><p id="">Urban land cover information extraction is a hot topic within urban studies. Heterogeneous spectra of high resolution imagery&mdash;caused by the inner complexity of urban areas&mdash;make it difficult. In this paper a hierarchical object oriented classification method over an urban area is presented. Combining QuickBird imagery and light detection and ranging (LIDAR) data, nine kinds of land cover objects were extracted. The Spectral Shape Index (SSI) method is used to distinguish water and shadow from black body mask, with 100% classification accuracy for water and 95.56% for shadow. Vegetation was extracted by using a Normalized Difference Vegetation Index (NDVI) image at first, and then a more accurate classification result of shrub and grassland is obtained by integrating the height information from LIDAR data. The classification accuracy of shrub was improved from 85.25% to 92.09% and from 82.86% to 97.06% for grassland. More granularity of this classification can be obtained by using this method. High buildings and low buildings can, for example, be distinguished from the original building class. Road class can also be further classified into roads and crossroads. The comparison of the classification accuracy between this method and the traditional pixel-based method indicates that the total accuracy is improved from 69.12% to 89.40%.</p>

DOI

[4]
Cleve C, Kelly M, Kearns F R, et al.Classification of the wildland-urban interface: a comparison of pixel-and object-based classifications using high-resolution aerial photography[J]. Computers, Environment and Urban Systems, 2008,32(4):317-326.

[5]
Michel J, Youssefi D, Grizonnet M.Stable mean-shift algorithm and its application to the segmentation of arbitrarily large remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015,53(2):952-964.Not Available

DOI

[6]
Wang M, Li R X.Segmentation of high spatial resolution remote sensing imagery based on hard-boundary constraint and two-stage merging[J]. IEEE Transactions on Geoscience and Remote Sensing, 2014,52(9):5712-5725.Not Available

DOI

[7]
Sziranyi T, Shadaydeh M.Segmentation of remote sensing images using similarity-measure-based fusion-MRF model[J]. IEEE Geoscience and Remote Sensing Letters, 2014,11(9):1544-1548.Not Available

DOI

[8]
Sammouda R, Touir A, Reyad Y A, et al.Adapting artificial hopfield neural network for agriculture satellite image segmentation[C]. Proceedings of the International Conference on Computer Applications Technology, 2013.

[9]
邓富亮,唐娉,刘源,等.引入松弛因子的高分辨率遥感影像自动多层次分割[J].遥感学报,2013,17(6):1492-1507.针对当前高分辨率遥感影像多层次分割尺度参数设置缺少理论框架支持、人为因素影响较多等缺点，提出一种引入松弛因子的高分辨率遥感影像自动多层次分割方法。该方法利用1个松弛因子调节引导区域对象合并的异质性值大小，通过控制每次递归合并区域的对象个数，提高了整体分割的速度；以区域对象间异质性平均值作为基数，引入另一个松弛因子控制分割过程中层次输出的尺度参数，使整个分割过程自动得到不同尺度的多层次分割结果。实验结果表明，该方法具有较高的分割质量，能够满足遥感影像分析及地物提取的精度要求，并且减少了人为因素影响，提高了自动化程度。但是，对于复杂图像内容的地物目标边界处理和减少狭长区域对象的出现还需要进一步深入研究和实践。

DOI

[ Deng F L, Tang P, Liu Y, et al.Automated hierarchical segmentation of high-resolution remote sensing imagery with introduced relaxation factors[J]. Journal of Remote Sensing, 2013,17(6):1492-1507. ]

[10]
沈占锋,骆剑承,胡晓东,等.高分辨率遥感影像多尺度均值漂移分割算法研究[J].武汉大学学报·信息科学版,2010,35(3):313-317.？根据高分辨率遥感影像信息提取过程中对影像的对象化分割的需求,分析了均值漂移分割算法的原理,并对其多尺度分割方法进行了设计与实现。实验证明该算法具有较好的影像分割精度。

[ Shen Z F, Luo J C, Hu X D, et al.A mean shift multi-scale segmentation for high-resolution remote sensing images[J]. Geomatics and Information Science of Wuhan University, 2010,35(3):313-317. ]

[11]
Liu J, Li P, Wang X.A new segmentation method for very high resolution imagery using spectral and morphological information[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015,101:145-162.Image segmentation is a key and prerequisite step for object-based analysis of very high resolution (VHR) imagery. Most existing image segmentation methods use either spectral or spatial information of an image alone. A novel image segmentation method for VHR multispectral images using combined spectral and morphological information is proposed in this paper. The method can be summarized as follows. First, a morphological derivative profile is calculated from an original multispectral image and combined with the spectral bands to quantify spectral-morphological characteristics of a pixel, which are considered as a criterion of homogeneity of neighboring pixels. Image segmentation is then conducted using a seeded region-growing procedure, which is based on the seed points automatically generated from the gradient image and dynamically added and the similarity between a seed pixel and its neighboring pixels in terms of spectral-morphological characteristics. The obtained segmentation result is further refined by a region merging procedure to generate a final segmentation result. The proposed method is evaluated using three VHR images of urban and suburban areas and compared with two existing segmentation methods, in terms of visual inspection, quantitative evaluation and indirect evaluation. Experimental results demonstrate that the joint use of spectral and morphological information outperformed the use of morphological information alone. Furthermore, the proposed image segmentation method performed better than existing methods. The proposed image segmentation method is well applicable to the segmentation of VHR imagery over urban and suburban areas.

DOI

[12]
Wang C, Shi A Y, Wang X, et al.A novel multi-scale segmentation algorithm for high resolution remote sensing images based on wavelet transform and improved JSEG algorithm[J]. Optik-International Journal for Light and Electron Optics, 2014,125(19):5588-5595.Considering the difficulties in image segmentation caused by the complexity of diverse ecological environments and various artificial targets in high resolution remote sensing images, especially in city scene, and in order to overcome the limitations existing in the traditional segmentation algorithm, JSEG (J-Segmentation), for high resolution remote sensing image segmentation and to further improve the segmentation accuracy, WJSEG (Wavelet-JSEG), a novel multi-scale segmentation algorithm based on wavelet transform, is proposed, which is an improved JSEG algorithm. WJSEG is an improved form of JSEG in relation to three aspects, including color quantization, multi-scale segmentation and region merging by introducing the multi-scale analysis tool based on wavelet transform. Experiments have been conducted on high resolution SPOT 5 pan-sharpened multispectral image and IKONOS panchromatic image. These experimental results were compared with those gained by the traditional JSEG algorithm and the famous commercial software named eCognition, which validated the effectiveness and reliability of the proposed WJSEG algorithm.

DOI

[13]
Zhang X, Xiao P, Feng X, et al.Hybrid region merging method for segmentation of high-resolution remote sensing images[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2014,98:19-28.ABSTRACT Image segmentation remains a challenging problem for object-based image analysis. In this paper, a hybrid region merging (HRM) method is proposed to segment high-resolution remote sensing images. HRM integrates the advantages of global-oriented and local-oriented region merging strategies into a unified framework. The globally most-similar pair of regions is used to determine the starting point of a growing region, which provides an elegant way to avoid the problem of starting point assignment and to enhance the optimization ability for local-oriented region merging. During the region growing procedure, the merging iterations are constrained within the local vicinity, so that the segmentation is accelerated and can reflect the local context, as compared with the global-oriented method. A set of high-resolution remote sensing images is used to test the effectiveness of the HRM method, and three region-based remote sensing image segmentation methods are adopted for comparison, including the hierarchical stepwise optimization (HSWO) method, the local-mutual best region merging (LMM) method, and the multiresolution segmentation (MRS) method embedded in eCognition Developer software. Both the supervised evaluation and visual assessment show that HRM performs better than HSWO and LMM by combining both their advantages. The segmentation results of HRM and MRS are visually comparable, but HRM can describe objects as single regions better than MRS, and the supervised and unsupervised evaluation results further prove the superiority of HRM.

DOI

[14]
邓富亮,杨崇俊,曹春香,等.高分辨率影像分割的分形网络演化改进方法[J].地球信息科学学报,2014,16(1):95-101.分形网络演化是针对高分辨遥感影像的高精度分割方法。它是以像元自下而上进行地物域合并，直至满足区域对象间异质性值大于预设阈值，停止区域合并得到最终分割结果。当对大数据量遥感影像进行分割时，形成初始区域对象的速度较慢，并且数量较多，导致分割时间长，有待在整体分割效率上进一步提高。一种有效的改进措施是采用某种分割方法，快速生成初始区域对象，然后再以初始分割结果区域对象进行区域合并。本文提出一种自动种子点的并行区域生长分割方法，用于快速生成初始区域对象；提出均匀数据划分的并行区域生长策略及消除数据划分线两侧的区域对象方法；采用OpenMP并行技术实现并行区域生长过程。分割效果对比和效率分析结果表明，本文提出的初始分割方法效率较高，并且分割结果可重现，从可信度、通用性角度来看，具有较高的实用价值。

DOI

[ Deng F L, Yang C J, Cao C X, et al.An improved method of FNEA for high resolution remote sensing image segmentation[J]. Journal of Geo-Information Science, 2014,16(1):95-101. ]

[15]
eCognition Developer. eCognition developer 8.7: reference book[M]. Munich, Germany: Trimble Germany GmbH, 2011.

[16]
何敏,张文君,王卫红.面向对象的最优分割尺度计算模型[J].大地测量与地球动力学,2009,29(1):106-109.面向对象的影像分析方法适合高分辨率遥感影像信息提取，其核心问题在于实现对高分辨率遥感影像的多尺度分割。提出了一种针对面向对象分析方法中多尺度分割的最优分割尺度计算模型，并进行了影像分割实验。结果表明，此模型能快速地获取可靠的最优分割尺度。

[ He M, Zhang W J, Wang W H.Optimal segmentation scale model based on object-oriented analysis method[J]. Journal of Geodesy and Geodynamics, 2009,29(1):106-109. ]

[17]
Witharana C, Civco D L.Optimizing multi-resolution segmentation scale using empirical methods: exploring the sensitivity of the supervised discrepancy measure Euclidean distance 2 (ed2)[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2014,87:108-121.Multiresolution segmentation (MRS) has proven to be one of the most successful image segmentation algorithms in the geographic object-based image analysis (GEOBIA) framework. This algorithm is relatively complex and user-dependent; scale, shape, and compactness are the main parameters available to users for controlling the algorithm. Plurality of segmentation results is common because each parameter may take a range of values within its parameter space or different combinations of values among parameters. Finding optimal parameter values through a trial-and-error process is commonly practiced at the expense of time and labor, thus, several alternative supervised and unsupervised methods for supervised automatic parameter setting have been proposed and tested. In the case of supervised empirical assessments, discrepancy measures are employed for computing measures of dissimilarity between a reference polygon and an image object candidate. Evidently the reliability of the optimal-parameter prediction heavily relies on the sensitivity of the segmentation quality metric. The idea behind pursuing optimal parameter setting is that, for instance, a given scale setting provides image object candidates different from the other scale setting; thus, by design the supervised quality metric should capture this difference. In this exploratory study, we selected the Euclidean distance 2 (ED2) metric, a recently proposed supervised metric, whose main design goal is to optimize the geometrical discrepancy (potential segmentation error (PSE)) and arithmetic discrepancy between image objects and reference polygons (number-of segmentation ratio (NSR)) in two dimensional Euclidean space, as a candidate to investigate the validity and efficacy of empirical discrepancy measures for finding the optimal scale parameter setting of the MRS algorithm. We chose test image scenes from four different space-borne sensors with varying spatial resolutions and scene contents and systematically segmented them using the MRS algorithm at a series of parameter settings. The discriminative capacity of the ED2 metric across different scales groups was tested using non-parametric statistical methods. Our results showed that the ED2 metric significantly discriminates the quality of image object candidates at smaller scale values but it loses the sensitivity at larger scale values. This questions the meaningfulness of the ED2 metric in the MRS algorithm鈥檚 parameter optimization. Our contention is that the ED2 metric provides some notion of the optimal scale parameter at the expense of time. In this respect, especially in operational-level image processing, it is worth to re-think the trade-off between execution time of the processor-intensive MRS algorithm at series of parameter settings targeting a less-sensitive quality metric and an expert-lead trial-and-error approach.

DOI

[18]
刘大伟,黄磊,李斌兵.一种基于对象相似性的遥感影像最优分割尺度评价方法[J].大地测量与地球动力学,2013,33(6):137-140.<p>针对遥感影像分割中最优尺度的选择问题，提出了根据参考对象和分割对象的灰度相似性和形状相似性构建分割尺度评价函数的方法。首先通过实地调绘和目视解译的方法选取典型地物作为参考对象，然后计算参考对象与分割对象的灰度相似性和形状相似性，并最终得到评价函数值，从而确定最优分割尺度。基于该评价函数，对World View2多光谱影像进行了实验，通过实验验证了该方法的有效性和适用性。</p>

[ Liu D W, Huang L, Li B B.An assessment method for optimal segmentation scale of remote sensing image based on similarity between objects[J]. Journal of Geodesy and Geodynamics, 2013,33(6):137-140. ]

[19]
陈春雷,武刚.面向对象的遥感影像最优分割尺度评价[J].遥感技术与应用,2011,26(1):96-102.<p>遥感影像分割决定了后续分类的精度,鉴于目前分割技术评价的研究缺乏且局限于主观判断的现状,以定量方法确定最优分割尺度。利用Definiens平台面向对象的分割算法,将组成对象的像素灰度值的标准差作为衡量对象内同质性的标准,用与邻域的平均差分的绝对值作为对象间的异质性度量变量,同时考虑面积权重的影响；根据上述3个评价指标,在考虑多光谱影像的基础上,构造了平均分割评价指数；基于该评价指数,以优度实验法对QuickBird多光谱影像进行了研究,并确定了不同地物类型的最优分割尺度。最后,利用平均对象匹配指数对评价结果进行了验证,并对评价方法的可行性进行了探讨。</p>

[ Chen C L, Wu G.Evaluation of optimal segmentation scale with object-oriented method in remote sensing[J]. Remote Sensing Technology and Application, 2011,26(1):96-102. ]

[20]
李秦,高锡章,张涛,等. 最优分割尺度下的多层次遥感地物分类实验分析[J].地球信息科学学报,2011,13(3):410-417.为了快速、准确地提取我国海岸带地区土地利用及其变化信息，选择 高分辨率遥感影像作为数据源，提出了最优分割尺度下的遥感多层次地物识别分类方法。首先，通过改进的局部方差法进行最优分割尺度的确定，建立影像中各对象 的方差均值与变化率随分割尺度变化曲线，确定方差均值的峰值，以变化率开始呈现下降趋势时所对应的分割值为最优分割尺度参考；然后，针对地物分类特征差异 选取各自适宜的分割尺度，建立多层次地物特征表达与规则，最后，实现最优尺度分割选择下的遥感多层次识别分类，即实现较大尺度下分割形成父对象，而较小尺 度下分割出其若干子对象的目标，提出了快速、自动化获取土地利用/覆盖图的策略流程。本文选取了广东省珠海市海岸带地区作为实验区，利用多层次遥感分类方 法进行地物识别分类。结果表明，其目视效果以及总体精度、Kappa系数，均优于传统方法和单一分割尺度下的影像分类方法。

DOI

[ Li Q, Gao X Z, Zhang T, et al.Optimal segmentation scale selection and evaluation for multi-layer image recognition and classification[J]. Journal of Geo-Information Science, 2011,13(3):410-417. ]

[21]
Drǎguţ L, Tiede D, Levick S R.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.ABSTRACT 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庐;. The degree of heterogeneity within an image-object is controlled by a subjective measure called the 'scale 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
Outlines

/

〈 〉