Techniques of classification for landuse/landcover with special reference to forest type mapping in Jaldapara Wildlife Sanctuary, Jalpaiguri District, West Bengal—a case study

The accuracy of three classification techniques namely Maximum likelihood, contextual and neural network for landuse/landcover with special emphasis on forest type mapping was evaluated in Jaldapara Wildlife Sanctuary area using IRS-1B LISS II data of Dec. 1994. The area was segregated into ten categories by using all the three classification techniques taking same set of training areas. The classification accuracy was evaluated from the error matrix of same set of training and validating pixels. The analysis showed that the neural net work achieved maximum accuracy of 95 percent, maximum likelihood algorithm with 91.06 percent and contextual classifier with 87.42 percent. It is concluded that the neural network classifier works better in heterogeneous and contextual in homogenous forestlands whereas the maximum likelihood is the best in both the conditions.     

Classifier ensembles for land cover mapping using multitemporal SAR imagery

SAR data are almost independent from weather conditions, and thus are well suited for mapping of seasonally changing variables such as land cover. In regard to recent and upcoming missions, multitemporal and multi-frequency approaches become even more attractive. In the present study, classifier ensembles (i.e., boosted decision tree and random forests) are applied to multi-temporal C-band SAR data, from different study sites and years. A detailed accuracy assessment shows that classifier ensembles, in particularly random forests, outperform standard approaches like a single decision tree and a conventional maximum likelihood classifier by more than 10% independently from the site and year. They reach up to almost 84% of overall accuracy in rural areas with large plots. Visual interpretation confirms the statistical accuracy assessment and reveals that also typical random noise is considerably reduced. In addition the results demonstrate that random forests are less sensitive to the number of training samples and perform well even with only a small number. Random forests are computationally highly efficient and are hence considered very well suited for land cover classifications of future multifrequency and multitemporal stacks of SAR imagery.     

Comparative performance of per-pixel classifiers using ers-1 sar data for classification of rice crop

Three-date ERS-1 SAR data acquired on August 24, September 28 and November 2, 1995, was used to classify rice crop in a predominant rice growing region of West Bengal. India, Artificial neural network, maximum likelihood, decision rute and K-Means clustering classifiers were used. Classification accuracy was evaluated from the error matrix of same set of training and validating pixels. Rice classification accuracy improved significantly using neural network classifier. The decision rule based classifier performed equally good for most of the sites, indicating the feasibility of deriving a common rule based algorithm for large area application. Law aecuracy was observed for maximum likelihood classifier.     

Evaluating effects of spectral training data distribution on continuous field mapping performance

Continuous field mapping has to address two conflicting remote sensing requirements when collecting training data. On one hand, continuous field mapping trains fractional land cover and thus favours mixed training pixels. On the other hand, the spectral signature has to be preferably distinct and thus favours pure training pixels. The aim of this study was to evaluate the sensitivity of training data distribution along fractional and spectral gradients on the resulting mapping performance.We derived four continuous fields (tree, shrubherb, bare, water) from aerial photographs as response variables and processed corresponding spectral signatures from multitemporal Landsat 5 TM data as explanatory variables. Subsequent controlled experiments along fractional cover gradients were then based on generalised linear models.Resulting fractional and spectral distribution differed between single continuous fields, but could be satisfactorily trained and mapped. Pixels with fractional or without respective cover were much more critical than pure full cover pixels. Error distribution of continuous field models was non-uniform with respect to horizontal and vertical spatial distribution of target fields. We conclude that a sampling for continuous field training data should be based on extent and densities in the fractional and spectral, rather than the real spatial space. Consequently, adequate training plots are most probably not systematically distributed in the real spatial space, but cover the gradient and covariate structure of the fractional and spectral space well.     

Soft classification for sub-pixel land cover extraction

With increasing resolution of the remotely sensed data the problems of images contaminated by mixed pixels arc frequent. Conventional classification techniques often produce erroneous results when applied to images dominated by mixed pixels. This may load to unrealistic representation of land cover, thereby, affecting efficient planning, management and monitoring of natural resources. Consequently, soft classification techniques providing sub-pixel land cover information may have to be utilised. From a range of soft classification techniques, the present study focuses on the utility of conventional maximum likelihood classifier and linear mixture modelling for sub-pixel. land cover classifications. The accuracy of the soft classifications has been assessed using distance measures and correlation co-efficient. The results show that linear mixture modelling has produced accuracies comparable to maximum likelihood classifier. Besides this the correlations between actual land cover proportions and proportions from linear mixture modelling, though not strong, arc statistically significant at 95% level of confidence. It has also been observed that the normalised likelihoods of maximum likelihood classifier also show strong correlations with the actual land cover proportions on ground and therefore has the potential to be used as a soft classification technique.     

多波段遥感数据的自组织神经网络降维分类研究

介绍了基于聚类分析的自组织特征映射神经网络分类方法，神经网络的输出层结构选用了3D结构，可以更好地保持多波段遥感数据中的内在拓扑结构；并选择天津大港地区的AsTER数据中的9个波段作为试验数据，通过对验证点的统计，分类精度达到了94％以上。     

ALCM: Automatic land cover mapping

It may be quite important for resource management people to extract single land cover class, at sub-pixel level from multi-spectral remote sensing images of different areas in single step processing. It has been observed, that neural network can be trained to extract single land cover class from multi-spectral remote sensing images, but they have problems in setting various parameters and slow during training stage. This paper present single land cover class water, extraction from mixed pixels present in multiple multi-spectral remote sensing data sets of same bands of AWiFS sensor of Resoursesat-1 (IRS-P6) satellite from different areas. In this work fuzzy logic-based algorithm, which is independent of statistical distribution assumption of data, has been studied at sub-pixel level to handle mixed pixels. It has been found; possibilistic c-means (PCM) algorithm takes the possibilistic view, that the membership of a feature vector in a class has nothing to do with its membership in other classes. Due to this, it was observed that PCM can extract only one class, from remote sensing multi-spectral data and it has produced 93.7% and 97.1% overall sub-pixel classification accuracy for two different data sets of different places using LISS-III (IRS-P6) reference data of same dates as of AWiFS data.     

Land-use classification using ASTER data and self-organized neutral networks

Operationally AVHRR and TM/TM+ data were used and a supervised maximum likelihood classification (MLH) was applied to depict land use changes in Beijing, providing basic maps for planning and development. With rapid growth of the city these are helpful to deal with higher resolution data, whereas new classification algorithms produce land use maps more accurate. In the paper, new sensor ASTER data and the Kohonen self-organized neural network feature map (KSOM) were tested.The TSOM classified 7% more accurately than the maximum likelihood algorithm in general, and 50% more accurately for the classes ‘residential area’ and ‘roads’. The results suggest that ASTER data and the Kohonen self-organized neural network classification can be used as an alternative data and method in a land use update operational system.     

变端元混合像元分解冬小麦种植面积测量方法

针对线性混合像元分解(Linear Spectral Unmixing,LSU)在端元(Endmember)个数不变情况下常会出现端元分解过剩现象导致分解结果精度不高的问题,以地物分布的聚集性特征为基础,提出了基于格网的变端元线性混合像元分解(Dynamic Endmember LSU,DELSU)方法.以冬小麦为研究...     

应用MODIS数据对陕西地区土地利用/覆盖变化的监测

应用MODIS数据监测陕西地区土地利用/覆盖变化。主要内容是利用陕西省MODIS影像辅助以ETM＋等数据进行最大似然法监督分类,根据分类的结果得到各个土地利用类型面积,然后与统计资料对比,进行土地利用/土地覆盖动态监测分析。     

遥感图像土地覆盖分类中多源特征数据选择研究

多源特征数据可以提高遥感图像的分类精度,选择合适的特征数据十分重要。利用基尼指数对多尺度纹理信息、主成分变换前三分量、地形数据等特征进行选择,选出最佳特征子集。利用支持向量机、神经网络分类法、最大似然法分别对全部特征数据和最佳特征子集结合多光谱数据进行分类。实验结果表明:基尼指数可以有效地对多源特征数据进行选择,特征选择可以提高分类器效率,提高分类精度。     

A comparative study of supervised classifiers on a subscene in Junagadh district,Gujarat

This paper describes the results of a comparative study of five classifiers viz., maximum likelihood, modified maximum likelihood, minimum distance to mean. Fisher and min-max, for classifying a subscene of Junagadh district using Landsat Thematic Mapper (TM) data. The kappa coefficient of agreement (k) and per cent correctly classified pixels for training data are used as measures of overall performance. It is observed that maximum likelihood and modified maximum likelihood classifiers perform better than the other three classifiers for this data set. Band combinations (3, 4, S) and (2, 3, 4, S) perform better than the usual combination (1,2,3,4), possibly because of presence of middle infrared band (band 5) on a scene dominated by vegetation cover. The band combination (1, 2, 3, 4, 5, 7) performed the best.     

基于ASTER数据的决策树自动构建及分类研究

 在对ASTER原始9个波段数据进行各种变换处理的基础上，采用数量化指标平均可分性方法确定参与分类的最佳特征组合； 结合研究区8种主要地物类型训练数据集，分别采用最大似然法、BP神经网络法和基于See 5.0数据挖掘的决策树分类法进行分类，提取主要地物的空间分布专题信息。经过379个野外样点的验证，结果表明： 决策树算法分类性能最优，神经网络算法次之，最大似然法效果最差； 与ENVI 4.1、ERDAS 8.7提供的传统决策树建立及分类方法比较，基于数据挖掘工具See 5.0和Cart的决策树生成和分类方法具有客观、高效率、分类性能可靠和精度高等优点。     

长时间序列Radarsat图像的神经网络模拟及土地覆盖变化的快速检测

土地覆盖的短期时空变化模式研究,对土地覆盖的快速、动态监测具有重要意义,也是遥感研究的新热点。本文利用2000—2001年的时间序列Radarsat图像,采用功率谱分析方法,对土地覆盖的短期时—空变化的周期特征进行了分析,由此建立了基于时间序列影像分析的神经网络预测模型,从植被主要生长季节的时间序列雷达卫星影像获取训练样本,对研究区域的典型土地覆盖的短期动态变化过程进行了学习。学习后的模型能够利用多个时间序列的Radarsat影像对下一时刻的影像进行模拟,并进一步检测变化。在模拟结果基础上,定义相对变化距离函数和检测门限,对模拟影像及实际影像中的变化区域进行了检测。检测精度范围在66.67%(农村居民点)—91.67%(水体)之间,平均检测精度为81.66%。由于时间序列信号的引入,神经网络模型能够较好地获取土地覆盖的短期动态变化信息。     

Object-based urban land cover classification using rule inheritance over very high-resolution multisensor and multitemporal data

Very high spatial and temporal resolution remote sensing data facilitate mapping highly complex and diverse urban environments. This study analyzed and demonstrated the usefulness of combined high-resolution aerial digital images and elevation data, and its processing using object-based image analysis for mapping urban land covers and quantifying buildings. It is observed that mapping heterogeneous features across large urban areas is time consuming and challenging. This study presents and demonstrates an approach for formulating an optimal land cover classification rule set over small representative training urban area image, and its subsequent transfer to the multisensor, multitemporal images. The classification results over the training area showed an overall accuracy of 96%, and the application of rule set to different sensor images of other test areas resulted in reduced accuracies of 91% for the same sensor, 90% and 86% for the different sensors temporal data. The comparison of reference and classified buildings showed ±4% detection errors. Classification through a transferred rule set reduced the classification accuracy by about 5%–10%. However, the trade-off for this accuracy drop was about a 75% reduction in processing time for performing classification in the training area. The factors influencing the classification accuracies were mainly the shadow and temporal changes in the class characteristics.     

Classifiers vs. input variables—The drivers in image classification for land cover mapping

The study investigates the performance of image classifiers for landscape-scale land cover mapping and the relevance of ancillary data for the classification success in order to assess and to quantify the importance of these components in image classification. Specifically tested are the performance of maximum likelihood classification (MLC), artificial neural networks (ANN) and discriminant analysis (DA) based on Landsat7 ETM+ spectral data in combination with topographic measures and NDVI. ANN produced high accuracies of more than 75% also with limited input information, while MLC and DA produced comparable results only by incorporating ancillary data into the classification process. The superiority of ANN classification was less pronounced on the level of the single land cover classes. The use of ancillary data generally increased classification accuracy and showed a similar potential for increasing classification accuracy than the selection of the classifier. Therefore, a stronger focus on the development of appropriate and optimised sets of input variables is suggested. Also the definition and selection of land cover classes has shown to be crucial and not to be simply adaptable from existing land cover class schemes. A stronger research focus towards discriminating land cover classes by their typical spectral, topographic or seasonal properties is therefore suggested to advance image classification.     

Approaches for delineating landslide hazard areas using different training sites in an advanced artificial neural network model

The current paper presents landslide hazard analysis around the Cameron area, Malaysia, using advanced artificial neural networks with the help of Geographic Information System (GIS) and remote sensing techniques. Landslide locations were determined in the study area by interpretation of aerial photographs and from field investigations. Topographical and geological data as well as satellite images were collected, processed, and constructed into a spatial database using GIS and image processing. Ten factors were selected for landslide hazard including: 1) factors related to topography as slope, aspect, and curvature; 2) factors related to geology as lithology and distance from lineament; 3) factors related to drainage as distance from drainage; and 4) factors extracted from TM satellite images as land cover and the vegetation index value. An advanced artificial neural network model has been used to analyze these factors in order to establish the landslide hazard map. The back-propagation training method has been used for the selection of the five different random training sites in order to calculate the factor’s weight and then the landslide hazard indices were computed for each of the five hazard maps. Finally, the landslide hazard maps (five cases) were prepared using GIS tools. Results of the landslides hazard maps have been verified using landslide test locations that were not used during the training phase of the neural network. Our findings of verification results show an accuracy of 69%, 75%, 70%, 83% and 86% for training sites 1, 2, 3, 4 and 5 respectively. GIS data was used to efficiently analyze the large volume of data, and the artificial neural network proved to be an effective tool for landslide hazard analysis. The verification results showed sufficient agreement between the presumptive hazard map and the existing data on landslide areas.     

基于空间自相关BP神经网络的遥感影像亚像元定位

亚像元定位技术是一种获取地物在混合像元中分布信息的有效方法.提出一种基于空间自相关函数的遥感影像BP神经网络亚像元定位方法,与传统的BP神经网路亚像元定位方法相比,该方法利用空间自相关函数Moran's I 在亚像素级上对定位结果进行约束,其结果更符合空间相关性假设理论.试验结果表明,该方法优于传统BP神经网络亚像元定...     

Approaches for delineating landslide hazard areas using different training sites in an advanced artificial neural network model

The current paper presents landslide hazard analysis around the Cameron area, Malaysia, using advanced artificial neural networks with the help of Geographic Information System (GIS) and remote sensing techniques. Landslide locations were determined in the study area by interpretation of aerial photographs and from field investigations. Topographical and geological data as well as satellite images were collected, processed, and constructed into a spatial database using GIS and image processing. Ten factors were selected for landslide hazard including: 1) factors related to topography as slope, aspect, and curvature; 2) factors related to geology as lithology and distance from lineament; 3) factors related to drainage as distance from drainage; and 4) factors extracted from TM satellite images as land cover and the vegetation index value. An advanced artificial neural network model has been used to analyze these factors in order to establish the landslide hazard map. The back-propagation training method has been used for the selection of the five different random training sites in order to calculate the factor’s weight and then the landslide hazard indices were computed for each of the five hazard maps. Finally, the landslide hazard maps (five cases) were prepared using GIS tools. Results of the landslides hazard maps have been verified using landslide test locations that were not used during the training phase of the neural network. Our findings of verification results show an accuracy of 69%, 75%, 70%, 83% and 86% for training sites 1, 2, 3, 4 and 5 respectively. GIS data was used to efficiently analyze the large volume of data, and the artificial neural network proved to be an effective tool for landslide hazard analysis. The verification results showed sufficient agreement between the presumptive hazard map and the existing data on landslide areas.     

利用神经网络方法提取水稻种植面积—以湖北省双季早稻为例

在利用NOAA数据提取水稻种植面积的过程中，由于其地面分辨率较低，存在大量混合像元问题，使得提取精度不够理想，该文基于神经网络方法即可以提供多源数据的输入，又不受数据分布假设限制的特点，从NOAA图像演算最能反应朋稻分布信息的绿度指数（NDVI）和日夜温差值，将其重采样，然后加入对水稻生产区域有重要影响的土壤类型，土地利用类型及高程分布等信息，以TM图像作为准直值进行分类，获得较为理想的湖北省双季早稻种植面积。