基于不同特征的随机森林极化SAR图像分类

近些年,利用计算机对极化SAR图像进行分类逐渐成为遥感领域的一个研究热点.本文采用全极化SAR数据,利用不同的特征提取算法提取特征,并基于随机森林模型最终实现对江苏沿海滩涂的分类.首先采用H/α和Freeman两种分解算法提取极化特征参数,采用灰度共生矩阵提取纹理特征参数;然后将提取的所有特征进行不同的组合,构成不同的特征集;最后采用随机森林模型对不同特征集合进行分类和精度评估.结果表明仅用纹理特征对沿海滩涂进行分类时效果较差;利用极化分解提取出的散射特征进行分类的结果要优于矩阵元素特征的分类结果;综合了极化散射特征和纹理特征的组合方式在沿海滩涂的分类中可以取得最优的分类结果,总体精度和Kappa系数可以达到94.44%和0.9305,表明极化SAR图像中蕴含的不同方面的特征在分类中具有一定的互补性.     

基于样本加权PointNet++的输电通道点云分类研究

输电通道内地物要素复杂，机载LiDAR获取的电力线、杆塔、植被等地物点云密度差异大、空间分布不规则，实际应用中“所见即所得”的应用需求对点云的高效自动化分类带来挑战。将深度学习中的PointNet++算法用于输电通道机载点云自动分类研究，分析样本加权对不同密度点云数据分类精度的影响，利用两组实验数据验证算法的精度和效率，并与随机森林分类算法进行比较。结果表明：基于样本加权PointNet++的方法在输电通道点云自动化分类方面适用性更强，平均F1值87.14%，且分类精度和效率均优于随机森林方法。     

基于超像素统计量的随机森林遥感图像分类

针对遥感图像地物覆盖分类方法对图像空间分布信息利用不足的问题,提出一种基于超像素统计量的随机森林遥感图像分类方法。以北京市海淀区为研究区,选用Landsat-8卫星为主要数据源,通过改进SLIC超像素分割方法,使之适用于多光谱遥感图像中超像素的分割,提取超像素常见的六个统计量(最小值、最大值、均值、标准差、上四分位数、下四分位数)用于随机森林在遥感图像中的分类。实验结果表明,本文对研究区遥感图像的总体分类精度为89.01%,明显改善了对地物的错分和漏分现象,能够推广到Landsat-8遥感图像的地物覆盖分类工作中。     

利用可分性指数的极化SAR图像特征选择与多层SVM分类

可分性指数（SI）可用来选择各类地物的有效分类特征,但在多维特征以及地物可分性较好的情况下,只利用可分性指数进行特征选择不能有效去除特征之间的冗余性。基于此,提出了利用可分性指数并辅以顺序后退（SBS）算法进行特征选择与多层支持向量机（SVM）分类的方法。首先,由各类地物在所有特征下的可分性指数选择分类地物和特征;然后,以该地物的分类精度为评估依据,利用顺序后退法筛选特征;其次,由剩余地物之间的可分性指数和顺序后退法依次选择各类地物的分类特征;最后利用多层SVM进行分类。实验结果表明,与只利用可分性指数选择特征进行多层SVM分类的方法相比,所提方法的分类精度提高了2%,各类地物的分类精度均高于86%,且运行时间为原来方法的一半。     

基于WorldView-2的红树林种群分类——以深圳湾为例

本文以WorldView-2多光谱遥感影像数据为数据源,选取深圳湾地区的红树林自然保护区为研究区域,基于ENVI5.3选用神经网络、支持向量机和随机森林三种分类方法对该区域的红树林进行种群分类,并对分类结果进行了对比分析。结果表明:随机森林分类法总体精度为73.6842%Kappa系数为0.6780,优于其余两种分类法。     

基于资源三号影像的红树林物种分类研究

选取北部湾的东部地区(广东省和广西壮族自治区交界处)为研究区,探索利用资源三号卫星影像,建立大范围红树林遥感信息提取和精细分类模式。首先根据地物波谱特征提取水陆边线,并建立缓冲区生成红树林生长适宜区;再通过基于面向对象的阈值分类方法提取红树林植被信息;最后采用基于像元的最近邻、贝叶斯和随机森林方法对红树林进行树种级分类。结果表明:采用结合阈值法的面向对象遥感分类可以对整景资源三号卫星影像准确提取水陆边线并有效提取红树林分布区,采用基于像元的随机森林法对红树林树种级分类可以得到较好的分类效果(总体分类精度82.84%),优于最近邻法和贝叶斯法。基于混合方法的红树林提取与树种分类模式适用于大范围的红树林分类与制图,同时也证实了资源三号卫星数据应用于海岸带红树林研究的有效性。     

地物大小、对象尺度、影像分辨率的关系分析

遥感数据的分辨率越来越高, 给地物信息提取提出了新的挑战。利用基于像元的分类技术和基于多尺度分割的面向对象分类技术对高分辨率影像进行分类实验, 分析地物大小、对象尺度与影像分辨率的关系。实验结果表明不同地物由于其空间尺度不同, 与之相适宜的空间分辨率和对象尺度也不同, 在适宜分辨率的影像提取有较高的精度, 在适宜的对象尺度上提取对象信息有更高的精度。分析也表明面向对象的多尺度影像分类技术适应了不同地物有其相适宜的空间分辨率, 在适宜尺度影像层中提取地物, 其分类精度大大高于基于像元的分类方法。     

整合无人机和面向对象的农村居住环境信息提取

无人机遥感和面向对象图像分析技术在环境监测中得到越来越多的发展。然而在科学文献领域,使用无人机和面向对象制图农村居住环境的文献仍然很少。因此本研究构造一个整合框架用于提取农村居住环境中各类地物信息。首先利用尺度参数评估(ESP, Estimation of Scale Parameter)工具和专家判断来确定最优分割尺度参数;然后分别采用专家规则集和监督分类算法提取农村居住环境中各类地物;最后采用基于面的精度评价方法对分类性能进行评估。结果表明,利用ESP工具和专家判断确定最优分割尺度是可行的。总体精度为75.19%,说明基于规则的提取方法对研究区各类地物的提取效果不佳。但在农村居住环境中利用模板匹配结合阈值规则对太阳能热水器提取精度达92%。分析训练样本和特征对随机森林(RF,Random Forest)、支持向量机(SVM, Support Vector Machines)和K最近邻 (KNN, K-Nearest Neighbor) 分类器分类结果的影响,说明RF分类器对农村居住环境分类效果最好,总体分类精度高达91.34%。研究结果表明:该框架在农村居住环境地物提取方面是一种有价值的工具。     

随机森林中树的数量

随机森林是一种集成分类器,对影响随机森林性能的参数进行了分析,结果表明随机森林中树的数量对随机森林的性能影响至关重要。对树的数量的确定方法以及随机森林性能指标的评价方法进行了研究与总结。以分类精度为评价方法,利用UCI数据集对随机森林中决策树的数量与数据集的关系进行了实验分析,实验结果表明对于多数数据集,当树的数量为100时,就可以使分类精度达到要求。将随机森林和分类性能优越的支持向量机在精度方面进行了对比,实验结果表明随机森林的分类性能可以与支持向量机相媲美。     

基于关联规则的面向对象高分辨率影像分类

以北京市昌平区Geoeye-1高分辨率遥感影像为试验数据,研究了关联规则挖掘和面向对象相结合的地物分类方法。首先探讨了关联分类法的原理,再通过图像分割、特征提取、关联规则挖掘、分类器构建一系列过程实现了基于关联规则的面向对象高分辨率影像分类,最终评估分类精度并与K-近邻法进行了对比。结果表明,该方法具有较好精度,能够在一定程度上摆脱地物分类对于专家知识的依赖。     

Study on Mangrove Species Classification based on ZY-3 Image

A hybrid mangrove forest extraction and species classification model for large coastal region was proposed using a ZY-3 (ZiYuan-3) image in the eastern part of Beibu Gulf (located at the junction of Guangdong and Guangxi).Firstly,the coastline was extracted according to the spectral features of ZY-3 image.Secondly,the buffer zone along with the coastline was generated as the suitable area of mangrove distribution.Mangrove forests and non-mangrove areas were then further classified using threshold method based on object-based image classification in these areas.Finally,Mangrove forests were classified at specie level using three pixel-based supervised classification methods,k-Nearest Neighbor,Bayes,and Random Forest.The classification results and accuracies were also compared and discussed.The results indicated that object-based threshold method can extract the coastline accurately and map the mangrove forests effectively.The pixel-based random forest classifier can obtain satisfactory results (the overall accuracy of 82.24%) of mangrove species classification than the other classifiers.In summary,the hybrid mode proposed in this paper is suitable for mangrove forests mapping and species classification in a large region.It is also validated the feasibility application of ZY-3 image in coastal mangrove research.     

Submarine groundwater discharge in tidal flats revealed by space-borne synthetic aperture radar

Most space-borne sensors cannot detect subsurface features. Groundwater is a typical subsurface feature, and its discharge to coastal ocean waters plays an important role in transporting terrestrial chemical constituents and providing habitats for various species of fauna and flora. This is the first paper to report observational evidence for submarine groundwater discharge (SGD) in tidal flats using space-borne synthetic aperture radar (SAR). Tidal flats are composed of high-moisture-saturated sediments and water puddles. These shallow water puddles were imaged effectively by using SAR systems. The presence of water puddles is usually indicated by low radar backscatter in SAR images due to specular reflections on the water surface. This effect was proved by comparing radar backscattering coefficients obtained from two space-borne SAR systems, TerraSAR-X and RADARSAT-2, with those obtained from two theoretical scattering models, IEM and Oh model. We observed relatively large, widely distributed water puddles in belt shape along the upper parts of the tidal flat, which were confirmed to be related to the discharge of groundwater. The results of this research suggest that SAR can be a powerful tool for observing and determining the areal distributions of possible groundwater discharge in large tidal flats, which is normally difficult to detect with traditional measurement tools or survey techniques for groundwater discharge. We firmly believe that this technique can reduce significantly the efforts of field work to confirm SGD in tidal flats.     

Classification of sediments on exposed tidal flats in the German Bight using multi-frequency radar data

We present a new method for the extraction of roughness parameters of sand ripples on exposed tidal flats from multi-frequency synthetic aperture radar (SAR) data. The method is based on the Integral Equation Model (IEM) which predicts the normalized radar cross-section (NRCS) of randomly rough dielectric surfaces. The data used for this analysis were acquired in the German Bight of the North Sea by the Spaceborne Imaging Radar-C/X-Band SAR (SIR-C/X-SAR) in 1994. In-situ measurements of the root-mean-squared (rms) height and the correlation length of the sand ripples clearly demonstrate a relationship between these roughness parameters and the C-band NRCS determined from an ERS SAR image. Using the IEM we have calculated NRCS isolines for the three frequency bands deployed by SIR-C/X-SAR (L, C, and X band), as a function of the rms height and the correlation length of the sand ripples. For each SIR-C/X-SAR image pixel these two roughness parameters were determined from the intersections of the NRCS isolines at different radar bands, and they were used for a crude sediment classification for a small test area at the German North Sea coast. Comparing our results with available sediment maps, we conclude that the presented method is very promising for tidal flat classification by using data from presently existing airborne and future spaceborne multi-frequency SAR systems.     

基于无人机LiDAR点云的多类型植被覆盖滩涂地形滤波

滩涂是宝贵的自然资源,其地形的高精度反演具有重要科学价值,而现有技术和方法存在较大的局限性。无人机激光雷达技术可快速获取大范围滩涂高精度、高密度的三维点云数据,是滩涂地形反演的重要技术之一。如何对点云数据中滩涂植被进行高精度滤波是地形反演要解决的技术难点,尤其是当滩涂覆盖有茂密的异质性植被（如不同种类和几何形态）时,对滤波算法的通用性和鲁棒性提出了更高的要求。以上海崇明西滩湿地某一滩涂作为研究区域,选取草丛、灌木和高大乔木3类典型植被覆盖的局部区域,利用基于坡度滤波、渐进数学形态学滤波和布料模拟滤波3种常用的点云滤波算法进行点云数据处理,比较分析了3种方法的适用性。结果表明：布料模拟滤波对于3个典型区域实验结果的总误差分别为1.57%、0.16%和0.23%, Kappa系数分别为96.74%、98.70%和99.30%。相较于其他两种算法,布料模拟滤波精度更高,更适用于多类型植被覆盖滩涂区域。因此,采用布料模拟滤波对整个研究区域进行处理,取得了较好的滤波效果,与真实地面吻合度较高。最后,通过克里金插值得到整个研究区域高精度的地形数据。     

Comparing the effects of classification techniques on landscape-level assessments: pixel-based versus object-based classification

Landscape-level assessments, particularly the quantification of forest fragmentation, often involve calculating landscape metrics from classified remotely sensed images. The utility of these derived metrics is often assumed to be dependent on the quality of the classified images. We compared conventional, pixel-based classification and a newer method of object-based classification to determine the effects of these two methods on fragmentation analysis of Cockpit Country, Jamaica, West Indies. Both methods showed similar trends in fragmentation metrics; however, there were significant differences between the methods for the metrics that quantified landscape configuration. The object-based classification allowed for the easy inclusion of roads into the analysis, which produced more accurate maps that showed a significant difference in the size of the largest forest patch. The object-based method also allowed classification of forests to show the location and extent of core forest areas; we were therefore able to identify an area of core forest that had remained consistent over the study period as a significant area for conservation focus. We recommend that the object-based method be the method chosen for landscape analyses, particularly forest-fragmentation studies.     

Coastal geomorphological and land‐use and land‐cover study of Sagar Island,Bay of Bengal (India) using remotely sensed data

Morphological changes on Sagar Island are occurring at an alarming rate due to both natural and anthropogenic activities. The eastern part of the island is rapidly eroding due to destabilization and growth of tidal flats in the Muriganga estuary and the gradual shifting of water current towards the island. Over the last four years (1996–1999), the rate of coastal erosion has been much higher (11.35 km²) than accretion (2.65 km²), compared with the conditions prior to 1996. Coastal places like Dublat, Basantpur, Gobindapur, Collectorganj, and Sumatinagar have become the critical zones of erosion. The shorelines along the eastern and south‐western sides are receding. The extent of coverage of the paddy field, sandy beaches, and land vegetation has decreased from 1996 to 1999 by 15.7, 1.1, and 3.5 km², respectively. An integrated database of the island was generated using spatial and non‐spatial data collected through field survey, satellite images of IRS‐1C LISS III, and topomaps. Spatial data include coastal geomorphological landforms, land‐use and land cover, shoreline change, sandy beaches, coastal erosion sites, agricultural fields, aquaculture sites, and coastal riparian vegetations. Non‐spatial data include the demography and evolution of the island. The main critical environmental issues of the island are: (1) degradation of mangrove forests and coastal erosion; (2) overpopulation and over‐exploitation of living resources; and (3) destruction of seawalls. Further degradation may lead to extinction of a variety of species and scarcity of marine food unless properly managed and regulated.     

基于ALOS影像的盐城海滨湿地遥感信息分类方法研究

盐城海滨湿地类型丰富多样,湿地植物覆被类型之间的生态交错带十分明显,如何更为准确地获得海滨湿地覆盖信息,对湿地研究具有重要价值和意义。以ALOS影像为数据源,江苏盐城海滨湿地核心区为试验区,开展湿地信息遥感分类研究。在对研究区进行非监督分类,分析其限制分类精度原因基础上,针对研究区域的特点提出适合的分类精度改进方法。以非监督分类后的结果为模板,借助分区分层分类方法的思想,通过分析遥感影像光谱信息、纹理信息、主成分变换信息,得到知识规则,以基于知识规则修改的方法对芦苇、米草和盐蒿3种植被交错带进行修正。然后以基于GIS规则的方法对剩余区域进行修正。通过GPS数据进行精度检验,分类精度达到92.6829%,Kappa系数为0.9098。实验证明基于GIS规则和知识规则的分区分层分类法是提高海滨湿地遥感分类精度的有效方法。     

Annual seasonality in Sentinel-1 signal for forest mapping and forest type classification

ABSTRACT

The Sentinel-1 satellites provide the formerly unprecedented combination of high spatial and temporal resolution of dual polarization synthetic aperture radar data. The availability of dense time series enables the derivation and analysis of temporally filtered annual backscatter signals. The study concentrates on the use of Sentinel-1 seasonal backscatter signatures for forest area estimation and forest type classification. A classification method based on time series similarity measures is introduced and tested in three test areas covered by various forest types including broadleaf temperate, boreal and montane forests. The results are compared with two European-wide Copernicus high resolution layers, namely forest type and tree cover density (TCD). The correspondence of forest/non-forest maps and TCD is high in all test areas, with overall accuracies for forest/non-forest classification between 86% and 91% and Pearson correlation coefficients for TCD between 0.68 and 0.74. The forest type classification (non-forest, coniferous and broadleaf forest classes) provides best results in temperate forests with an overall accuracy of 85%; in boreal forest, the accuracy decreases to only 65%. Generally, the method provides reliable results for forest area estimation, including regions where methods based on static parameters are often problematic (mountainous areas), and it enables forest type classification in temperate forests.     

Monitoring post-tsunami vegetation recovery in Phang-Nga province,Thailand, based on IKONOS imagery and field investigations – a contribution to the analysis of tsunami vulnerability of coastal ecosystems

A major tsunami in December 2004 devastated the coastal ecosystems along the Andaman Sea coast of Thailand. Since intact coastal ecosystems provide many important services for local communities at the Andaman Sea, it is crucial to investigate to what extent (in terms of percentage area and speed) the affected ecosystems were capable of recovering after the tsunami. Field measurements and multi-date IKONOS imagery were used to estimate the recovery and succession patterns of coastal vegetation types in the Phang-Nga province of Thailand, three years after the tsunami. Thus, this study contributes to a holistic understanding of the ecological vulnerability of the coastal area to tsunamis. A zone-based change detection approach is applied by comparing two change detection techniques: the first method involves the calculation of a recovery rate based on multi-temporal TNDVI (transformed normalized difference vegetation index) images (TNDVI approach), whereas the second approach is a combined approach of the change vector analysis (CVA). Although these two methods provide different types of information (quantitative for the TNDVI approach, qualitative for the CVA), they are comparable in terms of results and accuracies. The results reveal that recovery processes vary based on the type of the ecosystem and, furthermore, are strongly influenced by human activities. Grasslands, coconut plantations and the mixed vegetation cover could recover faster than the mangroves and casuarina forests. Among the forest ecosystems, recovery rates of casuarina forests were higher than for mangroves, but the recovery area was smaller. This study also discusses the potential and some limitations and inaccuracies of applying high-resolution optical imagery for assessing vegetation recovery at a local scale.     

Image classification methods applied to shoreline extraction on very high-resolution multispectral imagery

Comprehension of vulnerability to coastal erosion in dynamic coastal environments strongly depends on accurate and frequent detection of shoreline position. The monitoring of such environments could benefit from the semi-automatic shoreline delineation method, especially in terms of time, cost, and labour-intensiveness. This article explores the potential of using a semi-automatic approach in delineating a proxy-based shoreline by processing high-resolution multispectral WorldView-2 satellite imagery. We studied the potential and differences of basic and easily accessible standard classification methods for shoreline detection. In particular we explored the use of high spatial and spectral resolution satellite imagery for shoreline extraction. The case study was carried out on a 40 km coastal stretch facing the Northern Adriatic Sea (Italy) and belonging to the Municipality of Ravenna. In this area a frequent monitoring of shoreline position is required because of the extreme vulnerability to erosion phenomena that have resulted in a general trend of coastal retreat over recent decades. The wet/dry shorelines were delineated between the classes of wet and dry sand, resulting from different supervised (Parallelepiped, Gaussian Maximum Likelihood, Minimum-Distance-to-Means, and Mahalanobis distance) image classification techniques and the unsupervised Iterative Self-Organizing Data Analysis Technique (ISODATA). In order to assign reliability to outcomes, the extrapolated shorelines were compared to reference shorelines visually identified by an expert, by assessing the average mean distance between them. In addition, the correlation between offset rates and different types of coast was investigated to examine the influence of specific coastal features on shoreline extraction capability. The results highlighted a high level of compatibility. The average median distance between reference shorelines and those resulting from the classification methods was less than 5.6 m (Maximum likelihood), whereas a valuable distance of just 2.2 m was detected from ISODATA and Mahalanobis. Heterogeneous coastal stretches exhibited a larger offset between extracted and reference shorelines than the homogeneous ones. To finally evaluate the coastal evolution of the area, results from Mahalanobis classification were compared to a shoreline derived from airborne light detection and ranging (lidar) data. The fine spatial resolution provided by both methodologies allowed a detailed Digital Shoreline Analysis System (DSAS) comparison, detecting an erosive trend within a wide portion of the study area.