面向对象高分辨遥感影像分类研究

高空间分辨率遥感影像采用传统基于像元分类方法精度较低,本文通过分析高分辨遥感影像特征,采用面向对象的最近邻监督分类方法对QuickBird影像进行分类研究,首先对影像进行对象分割,然后将分割对象信息、形状特征与及上下文联系等特征构成特征空间进行最近邻监督分类,并与传统的基于像元最近邻分类方法分类进行比较分析,结果表明,本方法能够较好的识别高分辨率地物类型,总精度为92.19%,Kappa系数为0.8835,较好地改善分类效果,适合高分辨遥感影像分类。     

基于分层分析的ASTER 影像土地利用/ 覆盖遥感监测研究

以龙海市为实验区, 利用ASTER 遥感数据, 在研究区典型地物光谱特征系统分析的基础上,进行基于分层分类思想的地物分类提取方法研究。首先将影像划分为独立的子区( 水体、植被覆盖区和非植被覆盖区域) 以避免分类过程中光谱的互相影响; 然后在每个独立的子区基础上根据各类地物的不同光谱特征和空间特征, 对各类地物进行逐层掩模、分层提取。结果表明该方法优于传统的监督和非监督分类效果。     

基于卷积神经网络的光学遥感影像分析综述

光学遥感影像包含大量的地物信息，图像复杂性高，如何充分利用影像中的特征信息准确进行识别一直是该领域应用的一个难题。卷积神经网络具有对复杂特征进行选择和提取的优势，在遥感影像识别中有着优异的表现。介绍光学遥感影像的特点和经典的卷积神经网络及其在光学遥感影像中的研究实例，并基于遥感影像数据集进行了网络性能分析。从场景分类、目标检测和图像检索三大领域，详细综述常用的遥感影像数据集和研究进展，并作算法性能分析。最后给出基于卷积神经网络的光学遥感影像识别在未来的研究方向。     

CBERS-02B卫星CCD数据质量评价与植被分类应用潜力

针对CBERS-02BCCD影像,结合图像客观质量评价方法、目视解译、自动解译等方法,以广西三娘湾一带为研究区,对影像的植被识别能力进行了综合评价。与TM数据的对比分析表明,CCD1、2、3波段辐射精度稍偏低,第4波段表现出较好的性能,19．5m的空间分辨率比TM能更好地显示出地物的几何纹理特征。RGB假彩色合成图像的地物目视解译效果较好。通过非监督分类、监督分类和决策树分类等常规方法对地物分类及对典型植被识别效果表明,CBERS-02BCCD遥感影像能较好地区分出红树林、桉树林、松树林和灌丛／农田等典型植被类型。     

基于高分辨率影像数据的耕地信息提取应用研究

遥感影像地物信息丰富,使得直接通过影像获取耕地地块信息成为可能。本文介绍了以高分辨率遥感影像数据为基础图件,应用ERDAS、ARCGIS、eCogntion软件平台,通过非监督分类法、监督分类法和面向对象分类法分别对试验区域提取耕地信息的过程。试验结果表明,遥感技术是获取耕地信息提取的有效途径之一,监督分类进度为75．5％,面向对象法精度为90％。面向对象分类方法提取精度较高,效果最好。     

结合地籍数据的高密度城区面向对象遥感分类 

利用高分辨率遥感影像和GIS辅助数据,对高密度城区进行面向对象的土地利用覆被分类研究。使用NAIP高分辨率航空遥感影像,在多尺度影像分割的基础上,针对特定地物选择合适的影像分割参数。采用决策树方法建立高密度城市地区的分类规则,并结合该地区地籍图数据作为辅助数据,逐步进行高密度城市地区地物信息提取。利用辅助数据进行面向对象的遥感分类效果优于单纯依靠遥感影像进行的分类,且有效提取了道路和复杂的房屋等信息,得到了理想的分类结果,其总分类精度从常规面向对象方法的84.08%提高到89.79%。利用辅助数据进行遥感分类提高了高分辨率遥感影像的分类精度,说明了利用辅助数据进行遥感分类方法的有效性。     

应用高分辨率遥感影像提取作物种植面积

利用中低分辨率遥感影像提取作物分类种植面积的精度,往往难以满足农业遥感估产的需要。随着新型传感器的不断出现,应用高分辨率遥感影像高精度地提取作物分类面积日益成为发展趋势。由于高分辨率遥感影像提供的地物纹理、色调与形状等信息更加丰富,当前基于对象的地物识别分类方法仍不成熟,处理操作中人为干预过多,而且较为复杂,因此尝试以地面调查信息为辅助参量,采用常规基于像元的最大似然法监督分类方法,依据多尺度遥感影像信息提取的原理,分阶段地逐步提取作物种植面积,以此为农业遥感估产服务。     

基于遥感影像的军事阵地动态监测技术研究

针对部队快速机动作战的军事要求,提出基于高分辨率遥感影像的军用阵地动态监测方法。借助面向对象的多尺度分割技术将阵地影像分割为同质对象,以提取各个对象的特征;针对监督分类和非监督分类的弊端,提出通过一定的先验知识制定分类规则的方法对遥感影像进行地物识别,在此基础上定性和定量地输出变化检测结果。实验结果表明:利用基于对象影像分析方法具有较高的识别精度,能够有效监测军事阵地变化。     

LiDAR点云支持下地物精细分类的实现方法

在遥感数据分类中,获取精细的地物类别无疑能够传递更加丰富的信息量,进一步加深对遥感数据的理解和解译。在机载LiDAR点云高程数据的支持下,提出并实现了遥感影像上地物精细分类的方法。为保证高精度地同种地物再划分,综合考虑配准、辅助数据源、首次回波、点云密度及影像空间分辨率4种因素,并重点解决了点云密度与影像空间分辨率不匹配的问题,利用决策树显著地提高了影像上建筑物、植被的分类数量,使点云与影像联合分类的优势得到体现,达到了分类精度与地物类别数量相统一的目的。     

多尺度分割的高分辨率遥感影像分类对比研究

为进一步提高高分辨率影像地物分类精度,以高分2号卫星影像为研究数据,根据高分2号卫星遥感影像光谱信息以及高空间分辨的结构特点,从遥感影像数据分割的尺寸效果及其各种地物显著特征着手,通过局部方差法寻求出各类地物的最优分割尺度,并建立尺度网络层,利用继承进行多尺寸下多特征的整体融合,在最优尺度层下根据光谱特征、形状特征对高分2号影像进行多特征融合与多尺度分割实验,并在此基础上进行了典型地物的分类对比研究。结果表明多特征融合多尺度分割能够较好利用高空谱信息提高地物分类精度。     

江苏西溪贝丘遗址的高光谱遥感考古研究

针对江苏西溪贝丘遗址,通过样品采集对干净贝壳、自然状况下贝壳、土壤样品和地表裸地进行了光谱测量和分析。研究发现贝壳具有明显的光谱特征吸收峰,中心波长为2288nm处,吸收峰带宽在2098~2335nm范围内,与背景土壤粘土2180nm处的特征吸收峰和大多数碳酸岩类地表矿物质2350nm附近的吸收峰具有很好的可区分性。结合贝壳2288nm处特征吸收峰,利用该贝丘遗址发掘前的空间分辨率为6.6m的128波段OMIS航空高光谱遥感飞行数据成功地对该贝丘遗址信息进行有效地提取和圈定。本研究为利用高光谱遥感进行贝丘遗址考古提供了理论依据和有效技术途径。     

Ground water potential in a semi-arid region of Andhra Pradesh - a geographical information system approach

Continuous and adequate supplies of potable water from ground reservoirs are important for sustained agriculture, industry and domestic use throughout huge semi-arid regions of India. The present paper describes an approach to investigating groundwater potential over extensive geographical areas and illustrates its potential with reference to watershed planning in the large Varaha River Basin (VRB), Andhra Pradesh, India. The method involves the creation of a systematic database of information from satellite data for reconnaissance survey before going for field exploration. Colour composite images from Landsat Thematic Mapper and Indian Remote Sensing (IRS) satellite were used to interpret various thematic maps of the Varaha river basin. SPOT 1 MLA data of band 3 on a 1:250 000 scale was used for improving the accuracy of interpretation of topographic units due to its higher resolution and stereo coverage. Slope and other coverages were derived from topographic maps. The thematic and topographic information was digitized and ERDAS Imagine GIS software was used to analyse this information. Groundwater potential zones were delineated through subjective weights assigned to interpreted thematic and derived topographic units according to their likely infiltration capacities. Seven categories of groundwater potential ranging from very good to poor were derived automatically. Field measurements were then made within a selection of these categories to check the groundwater potential at selected sites. The validity and effectiveness of using remote sensing and GIS techniques for improving the targeting of field observations for groundwater for a huge river basin is shown by comparing the inferred groundwater potential with the field measurements.     

陆地卫星遥感信息在区域地下水资源量计算中的应用

利用陆地卫星遥感信息对地下水资源量计算问题作了两方面的研究。首先，以辽宁省朝阳市为例，研究了在常规资料比较齐全的地区运用遥感技术结合常规方法计算区域地下水资源量的系统方法(简称RSNM概念性研究模型)。其次，以西藏年楚河流域为例，探讨了常规资料短缺地区运用遥感方法估算区域地下水资源量的可能性及其系统研究方法(简称DRSNM概念性研究模型)。RSNM模型和DRSNM模型可推广应用于其它地区的地下水定量研究。     

Satellite remote sensing of groundwater: Quantitative modelling and uncertainty reduction using 6S atmospheric simulations

Remote sensing has been successfully used in the exploration of natural resources such as groundwater. Satellite data with different spatial, spectral and temporal characteristics have been evaluated for their potential use in groundwater detection in arid and semi-arid regions. However, distortions and noises caused by the presence of the atmosphere in the radiometric wave transmission become serious impediments for quantitative analysis and measurement work. In the present study, oasis and desert ecotone (ODE), a nonlinear ecological transitional belt, in Qira, Xinjiang Uyghur Autonomous Region of China was selected for this research. The ODE boundary was defined on the basis of widely collected information from the study area, including environmental, sociological and economic data. A model of groundwater level distribution using remote sensing (GLDRS), which empirically relates satellite sensor spectral radiance with groundwater level, is developed via in situ measurement and field examination of soil moisture and groundwater. Next, the second simulation of the satellite signal in the solar spectrum (6S), a code enabling simulations of radiative transfer process on the Sun-target-sensor path, is used to reduce uncertainties in the calculation of groundwater level. Then, groundwater level is evaluated using 6S atmospheric corrected and uncorrected Landsat-7 Enhanced Thematic Mapper (ETM)+ images respectively along with isochronous meteorological information. Greater correspondence between field examined and satellite monitoring data is obtained from 6S atmospheric corrected image (correlation coefficient is 0.94) than from the uncorrected image (correlation coefficient is 0.83).     

Integrated remote sensing and GIS for groundwater exploration and identification of artificial recharge sites

IRS-LISS-II data along with other data sets have been utilized to extract information on the hydrogeomorphic features of a hard rock terrain in the Sironj area of Vidisha district of Madhya Pradesh, India. The study exhibits reservoir induced artificial groundwater recharge downstream of surface water reservoirs. IRS-LISS-II data have been supported by information derived from DEM, drainage and groundwater data analysed in a GIS framework. The present study attempts to select suitable sites for groundwater recharge in a hard rock area through recharge basins or reservoirs, using an integrated approach of remote sensing and GIS. Criteria for GIS analysis have been defined on the basis of groundwater conditions in the area and appropriate weightage has been assigned to each information layer according to its relative contribution towards the desired output. The integrated study helps in designing a suitable groundwater management plan for a hard rock terrain.     

Delineation of groundwater recharge sites using integrated remote sensing and GIS in Jammu district,India

With the increasing demands for water due to increasing population, urbanization and agricultural expansion, groundwater resources are gaining much attention, particularly in the Kandi region of Jammu district, which faces acute shortages of drinking water throughout the year. Groundwater development programmes need a large volume of data from various sources. This study used integrated remote sensing and geographic information system (GIS) techniques to provide an appropriate platform for convergent analysis of multidisciplinary data and decision making for artificial recharge to groundwater. Thematic maps were constructed using merged Linear Imaging Self Scanner (LISS)‐III and Panchromatic (PAN) remote sensing data and aquifer parameter thematic layers were prepared from conventional field data. The thematic layers of the aquifer parameters were integrated and a map showing the potential zones for artificial recharge to groundwater was generated. By superimposing the drainage network map over this artificial recharge zone map, and also considering the terrain conditions for artificial recharge, suitable sites for replenishing groundwater in the study area were identified.     

Water balance study and conjunctive water use planning in an irrigation canal command area: A remote sensing perspective

Water budgeting of the D-36 and D-36 A distributaries confined between Pedda Vagu, Korutla Vagu and Kakatiya main canal of the Sri RamSagar Project (SRSP) Command area was conducted using remote sensing derived crop areas, land cover information, irrigation tank inventory and source-wise distribution of irrigated areas, together with conventional meteorological, canal flows and well inventory data. A semi-empirical water balance model was developed and validated using remote sensing derived objective information of the command area and the validated model used for predicting the groundwater table under normal rainfall conditions. Recharge and water balance in the study area indicated that the net recharge to the aquifer is negative to the tune of 2.54 Mm³ resulting in a fall of the groundwater table by 0.79 m during 1992-93. However, normalized groundwater recharge and water balance estimates indicate an impending waterlogging problem with an annual groundwater table rise of 0.35 m. In view of existing water management practices, a conjunctive water use plan of rotational operation of aquifers and canals is suggested.     

基于GIS和RS的长清区南部山区水土流失研究

长清区南部山区是济南市地表水来源地和地下水补给区的重要组成部分,加强水源地水土流失的监测和治理,有利于保证济南市的用水安全。本文利用RUSLE模型,应用GIS和RS技术,对长清区南部山区进行了水土流失研究。研究发现本区水土流失程度较轻,水土流失严重区集中在山区;不同的土地利用类型水土流失程度不同,荒草地水土流失最为严重;最后根据本区存在的4种治理模式提出了不同的治理建议。     

高分辨率遥感影像阴影检测与补偿方法优化

针对现有阴影检测误检水体为阴影及阴影增强导致非阴影区域信息改变的问题,提出基于统计信息分析的主成分变换水体识别与剔除方法,并结合蓝光抑制算法和阴影同质区统计信息实现阴影补偿。首先使用归一化阴影指数(Normalized Difference Umbra Index,NDUI)实现阴影区域的初始检测,然后利用主成分变换后水体统计信息与阴影统计信息的差异实现水体的识别与去除,并结合小区域去除和数学形态学方法完成阴影区域的检测,最后使用蓝光抑制算法和阴影同质区的统计信息分别对H、I、S分量补偿,并将结果转换回RGB色彩空间完成阴影补偿。以不同地区的高分辨率WorldView2影像和UltraCam D影像对提出的方法进行测试,结果表明:该方法可有效解决水体误检测为阴影的问题,且能够降低阴影补偿对非阴影区域的影响。     

一种结合显著性检测和超像素分割的遥感信息提取算法研究*

针对目前显著性检测算法在复杂多目标遥感图像中检测能力不足问题,提出一种结合显著性检测和超像素分割的遥感信息提取算法。该算法首先通过Graph-based Visual Saliency(GBVS)方法检测出原始影像中部分显著性较高的区域,然后利用Simple Linear Iterative Clustering(SLIC)方法分割显著区域,并修正显著区域边缘得到训练样本数据,进一步对训练样本进行统计学习,构造显著目标提取的阈值区间,最后实现对整幅超像素图像的显著目标提取。实验结果表明,该算法具有较高的准确率和召回率,能更加有效地检测出遥感图像中的显著目标,比目前主流的显著区域检测算法提取效果更好,可以很好地应用于具有明显显著区域的复杂多目标遥感图像信息提取中。