Land cover classification with multi-source data using evidential reasoning approach

Land cover classification is the core of converting satellite imagery to available geographic data.However,spectral signatures do not always provide enough information in classification decisions.Thus,the application of multi-source data becomes necessary.This paper presents an evidential reasoning (ER) approach to incorporate Landsat TM imagery,altitude and slope data.Results show that multi-source data contribute to the classification accuracy achieved by the ER method,whereas play a negative role to that...     

Remote sensing classification of western Sierra Leone using landsat TM and ETM +

The study examines the changes of land cover/use resources for the period under investigation.An unsupervised vegetation classification is being performed that provides five distinctive classes and thus assesses these changes in five broad land cover classes-high/moist forests,forest regrowth,mixed savanna,bare land/ grass and water.The remote sensing images used in this work are both images of TM and ETM+in different time periods(1986 to 2001)to determine land cover/use changes.A fairly accuracy report is recorded after performing the unsupervised classification,which shows vegetation has been depleted for over the years.Changes created are mostly human and to a lesser extent environment.Human activities are mainly encroachment thus altering the landscape through activities such as population growth,agriculture,settlements,etc.and environment due to some perceive climatic changes.This vegetation classification highlights the importance to acquire and publish information about the country’s partial vegetation cover and vegetation change including vegetation maps and other basic vegetation influencing factors,leading to an understanding of its evolution for a period.     

集成夜间灯光数据与Landsat TM影像的不透水面自动提取方法研究

利用多源遥感数据提取不透水面信息是一个重要的研究方向。针对以往研究中多需要人工选取不透水面样本进行模型训练的问题,本文通过整合夜间灯光遥感与Landsat TM影像中的空间和光谱信息实现了不透水面覆盖范围（Impervious Surface Area,ISA）的自动提取。首先根据夜间灯光的分布来定位ISA聚集的城市区域的位置,分别在城市区域内部和外部自动提取可靠性高的ISA及非ISA样本,然后通过迭代分类提取城市区域的ISA,再以此为样本对城市区域外部进行分类,最后将分类结果整合完成整幅影像的ISA提取流程。应用本方法对美国雪城地区的DMSP/OLS夜间灯光影像上提取了84个城市区域,提取精度大于95%。从中分别选择高ISA密度和低ISA密度的2个城市区域作为ISA提取的测试区,本文方法在城市区域内的ISA提取总体精度与kappa系数分别为88.23%和0.63;在城市区域外部为78.6%和0.54,均优于人工样本选取方法的提取精度,表明该方法能够实现精度稳定且高效的ISA自动提取。     

Combining Landsat-8 spectral bands with ancillary variables for land cover classification in mountainous terrains of northern Pakistan

Landsat-8 spectral values have been used to map the earth's surface information for decades. However, forest types and other land-use/land-cover(LULC) in the mountain terrains exist on different altitudes and climatic conditions. Hence, spectral information alone cannot be sufficient to accurately classify the forest types and other LULC, especially in high mountain complex. In this study, the suitability of Landsat-8 spectral bands and ancillary variables to discriminate forest types, and other LULC, using random forest(RF) classification algorithm for the Hindu Kush mountain ranges of northern Pakistan, was discussed. After prior-examination(multicollinearity) of spectral bands and ancillary variables, three out of six spectral bands and five out of eight ancillary variables were selected with threshold correlation coefficients r20.7. The selected datasets were stepwise stacked together and six Input Datasets(ID) were created. The first ID-1 includes only the Surface Reflectance(SR) of spectral bands, and then in each ID, the extra one ancillary variable including Normalized Difference Vegetation Index(NDVI), Normalized Difference Water Index(NDWI), Normalized Difference Snow Index(NDSI), Land Surface Temperature(LST), and Digital Elevation Model(DEM) was added. We found an overall accuracy(OA) = 72.8% and kappa coefficient(KC) =61.9% for the classification of forest types, and other LULC classes by using the only SR bands of Landsat-8. The OA = 81.5% and KC=73.7% was improved by the addition of NDVI, NDWI, and NDSI to the spectral bands of Landsat-8. However, the addition of LST and DEM further increased the OA, and Kappa coefficient(KC) by 87.5% and 82.6%, respectively. This indicates that ancillary variables play an important role in the classification, especially in the mountain terrain, and should be adopted in addition to spectral bands. The output of the study will be useful for the protection and conservation, analysis, climate change research, and other mountains forest-related management information.     

Detecting Arctic snow and ice cover with FY-1D global data

Chinese meteorological satellite FY-1D can obtain global data from four spectral channels which include visible channel(0.58-0.68 μm) and infrared channels(0.84-0.89 μm,10.3-11.3 μm,11.5-12.5 μm).2366 snow and ice samples,2024 cloud samples,1602 land samples and 1648 water samples were selected randomly from Arctic imageries.Land and water can be detected by spectral features.Snow-ice and cloud can be classified by textural features.The classifier is Bayes classifier.By synthesizing five d ays classifying result of Arctic snow and ice cover area,complete Arctic snow and ice cover area can be obtained.The result agrees with NOAA/NESDIS IMS products up to 70%.     

Land use and land cover changes in Jahlma watershed of the Lahaul valley, cold desert region of the northwestern Himalaya, India

Land use and land cover changes that occurred during the period from 1991 to 2001 in the Jahlma watershed of the Lahaul valley, a cold desert region of the northwestern Himalaya, were evaluated using land use data and visual interpretation of IRS Satellite imageries. The results revealed that out of the six major land use forms within the watershed, land areas under agriculture, kitchen garden and settlement land were found increased, whereas a declining trend was recorded in areas under grassland, barren land and Salix plantation. The cultivated land within the watershed increased from 54.87 % （total of agriculture land, kitchen garden, grassland, barren land, Salix plantation and residential area） in 1991 to 56.89 % in 2OOl, corresponding to an expansion of 4.41 ha. On the other hand, the areas of grassland decreased from 31.41% in 1991 to 29.81% in 2001. Such a dramatic land use and land cover changes taking place within the 33 km^2 watershed area in a single decade deafly indicates the prevailing danger of land degradation and environmental deterioration in the region.     

Extending self-organizing maps for supervised classification of remotely sensed data

An extended self-organizing map for supervised classification is proposed in this paper. Unlike other traditional SOMs, the model has an input layer, a Kohonen layer, and an output layer. The number of neurons in the input layer depends on the dimensionality of input patterns. The number of neurons in the output layer equals the number of the desired classes. The number of neurons in the Kohonen layer may be a few to several thousands, which depends on the complexity of classification problems and the classification precision. Each training sample is expressed by a pair of vectors: an input vector and a class codebook vector. When a training sample is input into the model, Kohonens competitive learning rule is applied to selecting the winning neuron from the Kohonen layer and the weight coefficients connecting all the neurons in the input layer with both the winning neuron and its neighbors in the Kohonen layer are modified to be closer to the input vector, and those connecting all the neurons around the winning neuron within a certain diameter in the Kohonen layer with all the neurons in the output layer are adjusted to be closer to the class codebook vector. If the number of training samples is sufficiently large and the learning epochs iterate enough times, the model will be able to serve as a supervised classifier. The model has been tentatively applied to the supervised classification of multispectral remotely sensed data. The author compared the performances of the extended SOM and BPN in remotely sensed data classification. The investigation manifests that the extended SOM is feasible for supervised classification.     

亚像元制图适应性分析与评价——以天津市津南区和北京市海淀区土地覆被制图为例

亚像元制图作为一种降尺度分类方法,可利用低分辨率影像获取高分辨率分类图。本文旨在探讨亚像元制图的降尺度分类结果与高分辨率影像分类精度和分类特征上的一致性。实验以天津市津南区和北京市海淀区为研究区,分别对中空间分辨率影像（TM或HJ）进行亚像元制图和对高空间分辨率影像（ALOS或ZY）进行硬分类得到相同空间分辨率的分类结果,从绝对精度、相对精度、空间结构和空间格局上,对2幅分类结果进行分析和评价。实验结果显示：（1）分类精度上,TM和HJ影像的亚像元制图结果,以地面验证样本为参考的绝对总体精度分别为84%和82%,以高分辨率影像（ALOS和ZY影像）硬分类结果,为参考的相对总体精度分别为82%和77%;（2）分类特征上,中空间分辨率影像亚像元制图结果的空间相关性较强、斑块数量较少、聚集度较高,但与高分辨率影像分类结果的总体结构相似,各类别的面积比例基本一致。因此,亚像元制图结果在分类精度和分类特征上与高空间分辨率影像分类结果具有较强的一致性,在缺少高分辨率土地覆被制图时,可将亚像元制图获取的降尺度分类图作为替代数据。     

Fusion and classification of Beijing-1 small satellite remote sensing image for land cover monitoring in mining area

In order to promote the application of Beijing-1 small satellite(BJ-1) remote sensing data,the multispectral and panchromatic images captured by BJ-1 were used for land cover classification in Pangzhuang Coal Mining.An improved Intensity-Hue-Saturation(IHS) fusion algorithm is proposed to fuse panchromatic and multispectral images,in which intensity component and panchromatic image are combined using the weights determined by edge pixels in the panchromatic image identified by grey absolute correlation degree.This improved IHS fusion algorithm outper-forms traditional IHS fusion method to a certain extent,evidenced by its ability in preserving spectral information and enhancing spatial details.Dempster-Shafer(D-S) evidence theory was adopted to combine the outputs of three member classifiers to generate the final classification map with higher accuracy than that by any individual classifier.Based on this study,we conclude that Beijing-1 small satellite remote sensing images are useful to monitor and analyze land cover change and ecological environment degradation in mining areas,and the proposed fusion algorithms at data and decision levels can integrate the advantages of multi-resolution images and multiple classifiers,improve the overall accuracy and produce a more reliable land cover map.     

Extending self-organizing maps for supervised classification of remotely sensed data

An extended self-organizing map for supervised classification is proposed in this paper. Unlike other traditional SOMs, the model has an input layer, a Kohonen layer, and an output layer. The number of neurons in the input layer depends on the dimensionality of input patterns. The number of neurons in the output layer equals the number of the desired classes. The number of neurons in the Kohonen layer may be a few to several thousands, which depends on the complexity of classification problems and the classification precision. Each training sample is expressed by a pair of vectors ： an input vector and a class codebook vector. When a training sample is input into the model, Kohonen＇s competitive learning rule is applied to selecting the winning neuron from the Kohouen layer and the weight coefficients connecting all the neurons in the input layer with both the winning neuron and its neighbors in the Kohonen layer are modified to be closer to the input vector, and those connecting all the neurons around the winning neuron within a certain diameter in the Kohonen layer with all the neurons in the output layer are adjusted to be closer to the class codebook vector. If the number of training sam- ples is sufficiently large and the learning epochs iterate enough times, the model will be able to serve as a supervised classifier. The model has been tentatively applied to the supervised classification of multispectral remotely sensed data. The author compared the performances of the extended SOM and BPN in remotely sensed data classification. The investigation manifests that the extended SOM is feasible for supervised classification.     

Spatial distribution of land cover and vegetation activity along topographic gradient in an arid river valley, SW China

Anthropogenic activities have become more and more important in characterizing the landscape, but their impacts are still restricted by natural environments. This paper discusses the interactions of anthropogenic activity, vegetation activity and topography through describing the spatial distribution of land cover and vegetation activity （represented by Normalized Difference Vegetation Index, NDVI） along topographic gradient in a mountainous area of southwestern China. Our results indicate that the existing landscape pattern is controlled by anthropogenic activities as well as topographic factors. Intensive anthropogenic activities mainly occur in areas with relatively low elevation, gentle and concave slopes, as these areas are easy and convenient to attain for human. Because of the destruction by human, some land cover types （mainly grassland and shrub） are only found in relatively harsher environments. This study also finds that topographic wetness index （W） used in other places only reflects runoff generation capacity, but not indicate the real spatial pattern of soil water content in this area. The relationships between NDVI and W, and NDVI and length slope factor （LSF） show that runoff and erosion have complex effects on vegetation activity. Greater values of W and LSF will lead to stronger capacity to produce runoff and transport sediment, and thereby increase soil water content and soil deposition, whereas beyond a certain threshold runoff and erosion are so strong that they would destruct vegetation growth. This study provides information needed to successfully restore native vegetation, improve land management, and promote sustainable development in mountainous areas, especially for developing regions.     

Impacts of land cover changes on runoff and sediment in the Cedar Creek Watershed,St. Joseph River,Indiana, United States

The relation between runoff and sediment and land cover is investigated in the Cedar Creek Watershed （CCW）, located in Northeastern Indiana, United States. The major land cover types in this watershed are cultivated land, woodland and pasture /Conservation Reserve Program （CRP）, which account for approximate 90 % of the total area in the region. Moreover, land use was changed tremendously from aooo to 9004, even without regarding the effect of the crop rotation system （corn ＆ soybean）. At least 49 % of land cover types were changed into other types in this period. The land cover types, ranking by changing area from high to low series, are rye, soybean, corn, woodland and pasture/CRP. The CCW is divided into 21 subwatersheds, and soil and water loss in each sub-watershed is computed by using Soil and Water Assessment Tool （SWAT）. The results indicate that the variations in runoff and sediment have positive relation to the area of crops （especially corn and soybean）; sediment is more sensitive to land cover changes than runoff; more heavy rainfall does not always mean more runoff because the combination of different land cover types always modify runoff coefficient; and rye, soybean and corn are the key land cover types, which affected the variation in runoff and sediment in the CCW.     

蒙古国30米分辨率土地覆盖产品研制与空间格局分析

作为蒙古高原的重要组成单元,蒙古国的土地覆盖格局与变化对于东北亚的资源、环境、生态及可持续发展具有重要意义。针对本区域缺乏高精度、现势性的土地覆盖数据产品的问题,本研究利用Landsat TM影像,采用面向对象的分类方法开展蒙古国土地覆盖遥感数据产品研制与分析。首先针对蒙古国景观格局特征,自主研究了适宜于蒙古国的土地覆盖分类体系,基于面向对象的遥感解译技术方法研究了蒙古国自然地物和人工地物要素的提取算法规则与阈值,建立了一套完整的面向蒙古国的土地覆盖遥感解译技术方案,在分景解译基础上获取了蒙古国2010年土地覆盖分类产品。经验证,数据集一级类分类精度为92.34%,二级类分类精度为80.24%。蒙古国土地覆盖类型以裸地、草地、森林为主,其中裸地的面积最大,占总面积的48.64%,其分布比较集中连片,主要分布在蒙古国南部和西部;草地面积次之,占总面积的42.85%,其分布具有明显的地域性,主要集中在蒙古国北部湿润地区和河流附近;林地最少,占总面积的6.63%,以蒙古国北部及西北部高山地区为主要生长区域。整体上蒙古国土地覆盖空间格局呈现明显的区域差异与地类过渡性,从南向北依次为裸地、荒漠草地、典型草地、森林类型,其中荒漠草地在中部形成一条明显的分界条带。     

随机森林方法支持的复杂地形区土地利用/土地覆被分类研究

随机森林方法目前已经成为遥感分类机器学习中一种有效方法,探索基于中等分辨率的Landsat卫星数据与随机森林方法相结合对复杂地形区长时间序列数据的获取及土地利用/土地覆被变化及模拟研究是非常有意义的。本文基于Landsat8OLI卫星多光谱数据,采用随机森林分类方法对青海省湟水流域复杂地形区土地利用类型进行了分类研究。针对复杂地形区域的情况,将研究区进行地理分区,根据每个分区的特点,选择相应的地形特征参数,并通过提取Landsat 8数据的光谱信息与纹理信息构建最优特征集,探索随机森林方法在复杂地形区土地利用分类的适用性。结果表明：使用Landsat8OLI数据进行随机森林分类,能较好地得到湟水流域复杂地形区域的土地利用类型结果;光谱、地形及纹理信息的结合在不同分区的表现结果不同。在脑山区光谱与地形信息结合能使随机森林分类效果最佳,总体精度达到91.33%,Kappa系数为0.886;而在浅山区与川水区综合考虑光谱、地形、纹理信息进行随机森林分类效果最佳,浅山区与川水区总体精度分别达到92.09%和87.85%,Kappa系数分别为0.902和0.859;利用随机森林算法进行优化选择纹理特征组合可以在保证分类精度的同时能够快速地提取土地利用类型信息,为复杂地形区土地利用类型的区分提供了实际可行的方法。     

村镇规划国土GIS数据的组织方法研究

村镇规划国土GIS数据是村镇规划国土信息系统的数据基础,因为其涉及规划、国土、建设等多个行业,数据来源多样、结构复杂,凶此不能简单地照搬城市领域的相关建设经验,而需对其数据组织方法进行深入的研究.本文首先分析了村镇规划国土GIS数据的特点,然后从数据分类编码、数据集成、异构数据融合和数据的专题化管理等几个方面对其进行了探讨研究.其对我国新农村建设中村镇信息的规范化组织有一定的借鉴作用.     

一种融合文本中地理位置和土地利用/覆被信息的野生动物活动细粒度定位方法

文本蕴含大量地理位置描述信息,有效融合地理关联信息以实现文本的精细定位是地理信息服务的难点。本文提出一种融合土地利用/覆被信息的描述地理位置的细粒度定位方法：在文本描述地理关联信息（地理位置实体、土地利用/覆被实体与空间关系）抽取、土地利用/覆被精细分类与地理位置粗粒度匹配定位的基础上,使用自然语言空间关系近似转换模型,确定地理位置的细粒度定位范围;基于土地利用/覆被实体及其周边精细分类信息,在该范围内检索匹配,确定地理位置的细粒度定位坐标。本文以野生亚洲象活动/肇事监测文本为例开展实验,并用匹配率与位置精度评价定位质量,结果表明：本文方法显著提升了地理位置的细粒度定位质量,实验精确匹配率（81.51%）、位置误差距离的均值（65.97 m）及其≤50 m的比例（70.50%）均优于国内主流在线地理编码与地名检索服务结合空间关系或其单独使用结果。该方法有助于完善地理位置定位方法体系、提升地理信息空间化质量,并可服务于野生动物活动/肇事监测预警等精细定位任务。