Land Cover Classification Method Research Articles

A Comparison of Multisensor Integration Methods for Land Cover Classification in the Brazilian Amazon

Many data fusion methods are available, but it is poorly understood which fusion method is suitable for integrating Landsat Thematic Mapper (TM) and radar data for land cover classification. This research explores the integration of Landsat TM and radar images (i.e., ALOS PALSAR L-band and RADARSAT-2 C-band) for land cover classification in a moist tropical region of the Brazilian Amazon. Different data fusion methods—principal component analysis (PCA), wavelet-merging technique (Wavelet), high-pass filter resolution-merging (HPF), and normalized multiplication (NMM)—were explored. Land cover classification was conducted with maximum likelihood classification based on different scenarios. This research indicates that individual radar data yield much poorer land cover classifications than TM data, and PALSAR L-band data perform relatively better than RADARSAT-2 C-band data. Compared to the TM data, the Wavelet multisensor fusion improved overall classification by 3.3%-5.7%, HPF performed similarly, but PCA and NMM reduced overall classification accuracy by 5.1%-6.1% and 7.6%-12.7%, respectively. Different polarization options, such as HH and HV, work similarly when used in data fusion. This research underscores the importance of selecting a suitable data fusion method that can preserve spectral fidelity while improving spatial resolution.

고해상도 위성영상을 이용한 도시유역의 소강우 유출해석

Abstract : In this research, enhanced land-cover classification methods using high-resolution satellite image (HRSI) and GIS in terms of practicality and accuracy was proposed. It aims for understanding non-point pollutant origin/loading, assessment the efficiency of rainfall storage/infiltration facilities and sounds water-environment management. The result of applying enhanced land-cover classification methods to the urban region verifies that roof and road area are including various vegetations such as roof garden, flower bed in the median strip and street tree. This accounts for 3% of total study area, and more importantly it was counted as impervious area by GIS alone or conventional indoor work. The feasibility of the method was assessed by applying to rainfall-runoff analysis for three weak rainfall in the range of 7.1-10.5 mm events in 2000, Chiba, Japan. A good agreement between simulated and observed runoff hydrograph was obtained. In comparison, the hydrograph simulated with land-use parameters by the detailed land-use information of 10m grid had an error between 31%~71%, while enhanced method showed 4% to 29%, and showed the improvement particularly for reproducing observed peak and recession flow rate of hydrograph in weak rainfall condition.

Object-oriented classification of high-resolution remote sensing imagery based on an improved colour structure code and a support vector machine

This paper presents a new object-oriented land cover classification method that integrates raster analysis and vector analysis. The method adopts an improved colour structure code (CSC) for segmentation and support vector machine (SVM) for classification using high resolution (HR) QuickBird data. It combines the advantages of digital image processing (efficient improved CSC segmentation), geographical information systems (GIS) (vector-based feature selection), and data mining (intelligent SVM classification) to interpret images from pixels to objects and thematic information. The improved CSC segmentation not only achieves robust and accurate results but also combines boundary information that the traditional CSC algorithm does not consider. The SVM used for classification has the advantages of solving sparse sampling, nonlinear, high-dimensional and global optimum problems, compared with other classifiers. The results demonstrate that the new object-oriented classification method significantly outperforms some other objected-oriented classification methods such as the objected-oriented method based on traditional CSC and SVM, and perfect classification results are obtained from the classification processing, including not only the classification method, but also preprocessing, sample selection and post-processing.

Testing Multivariate Adaptive Regression Splines (MARS) as a Method of Land Cover Classification of TERRA-ASTER Satellite Images

This work proposes a new method to classify multi-spectral satellite images based on multivariate adaptive regression splines (MARS) and compares this classification system with the more common parallelepiped and maximum likelihood (ML) methods. We apply the classification methods to the land cover classification of a test zone located in southwestern Spain. The basis of the MARS method and its associated procedures are explained in detail, and the area under the ROC curve (AUC) is compared for the three methods. The results show that the MARS method provides better results than the parallelepiped method in all cases, and it provides better results than the maximum likelihood method in 13 cases out of 17. These results demonstrate that the MARS method can be used in isolation or in combination with other methods to improve the accuracy of soil cover classification. The improvement is statistically significant according to the Wilcoxon signed rank test.

Object-based land cover classification of shaded areas in high spatial resolution imagery of urban areas: A comparison study

A significant proportion of high spatial resolution imagery in urban areas can be affected by shadows. Considerable research has been conducted to investigate shadow detection and removal in remotely sensed imagery. Few studies, however, have evaluated how applications of these shadow detection and restoration methods can help eliminate the shadow problem in land cover classification of high spatial resolution images in urban settings. This paper presents a comparison study of three methods for land cover classification of shaded areas from high spatial resolution imagery in an urban environment. Method 1 combines spectral information in shaded areas with spatial information for shadow classification. Method 2 applies a shadow restoration technique, the linear-correlation correction method to create a “shadow-free” image before the classification. Method 3 uses multisource data fusion to aid in classification of shadows. The results indicated that Method 3 achieved the best accuracy, with overall accuracy of 88%. It provides a significantly better means for shadow classification than the other two methods. The overall accuracy for Method 1 was 81.5%, slightly but not significantly higher than the 80.5% from Method 2. All of the three methods applied an object-based classification procedure, which was critical as it provides an effective way to address the problems of radiometric difference and spatial misregistration associated with multisource data fusion (Method 3), and to incorporate thematic spatial information (Method 1).

Object-Based Land Cover Classification Using High-Posting-Density LiDAR Data

This study introduces a method for object-based land cover classification based solely on the analysis of LiDAR-derived information—i.e., without the use of conventional optical imagery such as aerial photography or multispectral imagery. The method focuses on the relative information content from height, intensity, and shape of features found in the scene. Eight object-based metrics were used to classify the terrain into land cover information: mean height, standard deviation (STDEV) of height, height homogeneity, height contrast, height entropy, height correlation, mean intensity, and compactness. Using machine-learning decision trees, these metrics yielded land cover classification accuracies > 90%. A sensitivity analysis found that mean intensity was the key metric for differentiating between the grass and road/parking lot classes. Mean height was also a contributing discriminator for distinguishing features with different height information, such as between the building and grass classes. The shape- or texture-based metrics did not significantly improve the land cover classifications. The most important three metrics (i.e., mean height, STDEV height, and mean intensity) were sufficient to achieve classification accuracies > 90%.

Segment-based stem volume retrieval in boreal forests using multitemporal ERS-1/2 InSAR data

In this study, we demonstrate a method for the combined land cover classification and boreal forest stem volume retrieval using multitemporal interferometric synthetic aperture radar (InSAR) data. The method utilizes the mean InSAR coherence image that is segmented into quasihomogeneous segments, and the land cover classes of the segments are determined based on their multitemporal InSAR signatures. Continuous stem volume estimates for the forest segments are then produced by inverting a semi-empirical backscattering coherence model. A group of forest stands with known stem volumes are required as training areas for determining the values of the model parameters. The performance of the method was studied by estimating the stem volumes of 4176 forest segments using 134 training stands. The results were compared with stem volume estimates produced by ground-based measurements and the satellite-based operational National Forest Inventory (NFI) of Finland. The performance of the method in stem volume estimation for stands larger than 1.5 ha (average stand size 3.2 ha) achieves a correlation coefficient of 0.87 and a root mean square error (RMSE) of 54% in comparison with ground-based reference data. The results are better than the estimates of the operational NFI that saturate at around 200 m3/ha. Markedly higher accuracy is to be expected when applying the method to larger stands in a large-area forest inventory.

Land Cover Classification in the Poyang Lake Region, China, Using Landsat TM and JERS-1 Synthetic Aperture Radar Data

The Poyang Lake is the largest fresh water lake in China. As an internationally important wetlands, conservation of wild birds needs updated information on land use and land cover in the Poyang Lake region. This paper introduced a non-parametric knowledge-based classification method (decision tree classifier) for land cover classification in the Poyang Lake region. We merged optical sensor (Landsat 5 TM) image with Japanese Earth Resource Satellite-1(JERS-1) synthetic aperture radar (SAR) images. The overall accuracy of the classification result was about 82%, of which forest was classified with higher accuracy (over 87%) and less errors of omission and commission. Main classification errors came from the similar spectrum of different land cover classes in winter. The seasonal dynamics should be considered for selecting optical satellite images for classification when using this pixel-based classification algorithm. The results of this study suggests that the non-parametric decision tree classifier together with fusion of optical and SAR images is an efficient method for mapping complex landscapes with agriculture, wetlands and forests.

A comparison of four methods to map biomass from Landsat-TM and inventory data in western Newfoundland

Spatial measures of forest biomass are important to implement sustainable forest management, monitor global change, and model forest productivity. Several methods for estimating forest biomass by remote sensing have been developed, but their comparative advantages have not been evaluated for large areas in Canada. This study compares four methods to map forest biomass on an extended pilot region (20,000 km 2) located in western Newfoundland. The methods include: (i) Direct Radiometric Relationships (DRR), (ii) k-Nearest Neighbors ( k-NN), (iii) Land Cover Classification (LCC), and (iv) Biomass from Cluster Labeling Using Structure and Type (BioCLUST). The results of each method were evaluated using an independent set of ground survey plots and compared with a baseline biomass map generated from biomass tables applied to forest inventory stand maps. Considering the root mean square error (RMSE) assessed with the inventory plots, the DRR, k-NN, and BioCLUST methods provided similar results, with average RMSE values of 59, 59, and 58 t/ha, respectively. Bias values were lowest for the k-NN method followed by DRR, BioCLUST, and LCC (6, −8, 17, and 42 t/ha, respectively). Assessed with the baseline map, the BioCLUST method produced the lowest RMSE (41 t/ha) and bias (−4 t/ha) followed by the DRR and k-NN methods, with RMSE values of 47 and 54 t/ha and bias values of 9 and 23 t/ha, respectively. The method using biomass tables applied on the classified TM image (LCC) provided the greatest RMSE and bias, but may be suitable for applications that do not require a high level of precision. The BioCLUST and LCC methods provided practical advantages for the type of data sets available. Overall, the choice of a method rests on both the availability of data sets and the level of precision of the results required.

A Method of Land Cover Classification for China Based on NDVI-Ts Space

A Method of Land Cover Classification for China Based on NDVI-Ts Space

Large-area urban growth observations—a hierarchical kernel approach based on image texture

Geografisk Tidsskrift, Danish Journal of Geography 105(2):39–47, 2005 A multi-scale classification algorithm based entirely on computation of image texture in moving windows is presented and discussed. The rationale for using kernel-based methods as opposed to segment-based methods for texture-based land cover classification is discussed, as well as, the need for identifying strategies that may limit the effects of the inherent weaknesses of the kernel-based methods. Part of the proposed solution to this problem is to omit the search for an optimum kernel size and rather use different sizes concurrently followed by a weighted summary overlay process based on associated scores. The spatial growth pattern of the city of Accra, Ghana, is assessed over a 17 years period using the algorithm. Land-cover maps showing the extension of the new residential areas on the fringe of Accra are produced from Landsat ETM satellite imagery. These areas constitute the rapidly expanding frontier of urban Accra into its rural surroundings. The classification discriminates between new residential areas that have a distinct urban appearance and areas that are in a transition phase between rural and urban use. Finally, comparisons are made to a previous satellite-based study of land cover changes for the period 1985–1991.

Data fusion of hyperspectral and SAR images

A novel technique is proposed for data fusion of earth remote sensing. The method is developed for land cover classification based on fusion of remote sensing images of the same scene collected from multiple sources. It presents a framework for fusion of multisource remote sensing images, which consists of two algorithms, referred to as the greedy modular eigenspace (GME) and the feature scale uniformity transformation (FSUT). The GME method is designed to extract features by a simple and efficient GME feature module, while the FSUT is performed to fuse most correlated features from different data sources. Finally, an optimal positive Boolean function based multiclass classifier is further developed for classification. It utilizes the positive and negative sample learning ability of the minimum classification error criteria to improve classification accuracy. The performance of the proposed method is evaluated by fusing MODIS/ASTER airborne simulator (MASTER) images and the airborne synthetic aperture radar (SAR) images for land cover classification during the PacRim II campaign. Experimental results demonstrate that the proposed fusion approach is an effective method for land cover classification in earth remote sensing, and improves the precision of image classification significantly compared to conventional single source classification.

Comparison of Land-Cover Classification Methods in the Brazilian Amazon Basin

Four distinctly different classifiers were used to analyze multispectral data. Which of these classifiers is most suitable for a specific study area is not always clear. This paper provides a comparison of minimum-distance classifier (MDC), maximumlikelihood classifier (MLC), extraction and classification of homogeneous objects (ECHO), and decision-tree classifier based on linear spectral mixture analysis (DTC-LSMA). Each of the classifiers used both Landsat Thematic Mapper data and identical field-based training sample datasets in a western Brazilian Amazon study area. Seven land-cover classes— mature forest, advanced secondary succession, initial secondary succession, pasture lands, agricultural lands, bare lands, and water—were classified. Classification results indicate that the DTC-LSMA and ECHO classifiers were more accurate than were the MDC and MLC. The overall accuracy of the DTCLSMA approach was 86 percent with a 0.82 kappa coefficient and ECHO had an accuracy of 83 percent with a 0.79 kappa coefficient. The accuracy of the other classifiers ranged from 77 to 80 percent with kappa coefficients from 0.72 to 0.75.

An assessment of the effectiveness of decision tree methods for land cover classification

Choice of a classification algorithm is generally based upon a number of factors, among which are availability of software, ease of use, and performance, measured here by overall classification accuracy. The maximum likelihood (ML) procedure is, for many users, the algorithm of choice because of its ready availability and the fact that it does not require an extended training process. Artificial neural networks (ANNs) are now widely used by researchers, but their operational applications are hindered by the need for the user to specify the configuration of the network architecture and to provide values for a number of parameters, both of which affect performance. The ANN also requires an extended training phase. In the past few years, the use of decision trees (DTs) to classify remotely sensed data has increased. Proponents of the method claim that it has a number of advantages over the ML and ANN algorithms. The DT is computationally fast, make no statistical assumptions, and can handle data that are represented on different measurement scales. Software to implement DTs is readily available over the Internet. Pruning of DTs can make them smaller and more easily interpretable, while the use of boosting techniques can improve performance. In this study, separate test and training data sets from two different geographical areas and two different sensors—multispectral Landsat ETM+ and hyperspectral DAIS—are used to evaluate the performance of univariate and multivariate DTs for land cover classification. Factors considered are: the effects of variations in training data set size and of the dimensionality of the feature space, together with the impact of boosting, attribute selection measures, and pruning. The level of classification accuracy achieved by the DT is compared to results from back-propagating ANN and the ML classifiers. Our results indicate that the performance of the univariate DT is acceptably good in comparison with that of other classifiers, except with high-dimensional data. Classification accuracy increases linearly with training data set size to a limit of 300 pixels per class in this case. Multivariate DTs do not appear to perform better than univariate DTs. While boosting produces an increase in classification accuracy of between 3% and 6%, the use of attribute selection methods does not appear to be justified in terms of accuracy increases. However, neither the univariate DT nor the multivariate DT performed as well as the ANN or ML classifiers with high-dimensional data.

A method for object-oriented land cover classification combining Landsat TM data and aerial photographs

Object-oriented classification techniques based on image segmentation are gaining interest as methods for producing output maps directly storable into Geophysical Information System (GIS) databases. A limitation in efficiently applying image segmentation is often represented by the spatial resolution of the image. This contribution proposes a method for overcoming this problem, based on the integrated use of images of different resolution. A high-resolution black and white (b/w) orthophoto and a subscene of a Landsat Thematic Mapper (TM) image have been used to obtain an object-oriented classification of the land cover of a study area in northern Italy. The method consists of a sequential application of segmentation and classification techniques. First, the TM image was classified using the maximum likelihood classifier and additional empirical rules. Subsequently, the orthophoto was segmented by applying a region-based segmentation algorithm. Finally, the classification of the segmented images was performed using as a reference the TM image previously classified. The resulting land cover map was tested for accuracy and the results are dicusssed.

Using remote sensing and GIS to analyse land cover change and its impacts on regional sustainable development

Policy and decision making in the context of sustainable development requires rapid, effective and efficient access to and integration of appropriate current information from a wide range of sources and disciplines, including land cover dynamic information derived from remotely sensed data. The analysis of data from high spatial resolution satellite sensors has potential in land cover monitoring. In this paper, a post-classification method is used to detect land cover change from multi-temporal satellite data, and particular attention is given to the selection of an appropriate method for land cover classification. The use of a Geographic Information System (GIS) allows further spatial analysis of the data derived from remotely sensed images and analysis of the impact of land cover change on regional sustainable development. The satellite-derived data used in this study are Thematic Mapper (TM) data acquired by Landsat-5 on 14 May 1985, 20 May 1987, 26 April 1990 and 20 May 1993 in Ansan City, Korea. The results obtained show that land cover change in the west coastal zone of Korea has occurred in the past decade as a result of both natural forces and human activities, which has in turn impacted on the regional sustainable development. The results thus provide very useful information to local government for decision making and policy planning.

Evaluation of speckle filtering and texture analysis methods for land cover classification from SAR images

This paper describes a comparative evaluation of several speckle reduction and texture analysis techniques, with particular emphasis on their applicability to supervised land cover classification from SAR images. Issues related to suppression of speckle in a uniform area, preservation of edges, and texture preservation are pursued in these filters. Quality of texture features is measured by the relevancy, discriminative power and ease of computation of the features. The discriminative power of texture features is measured using the Jeffreys-Matusita distance and classification performance measured on a validation set independent from the classifier's training set. Classifiers investigated are maximum-likelihood, multi-layer perceptron (MLP) and radial basis function (RBF) neural networks. Classification accuracy is measured by KHAT statistic calculated from confusion matrices. Two SAR images of ERS-1 and E-SAR programme showing different land cover categories within the regions of Douala and Ngaoundere (Cameroon), and a bi-polarized Synthetic Aperture Radar (SAR) image from an agricultural station near the city of Altona (Canada) are used for analysis. Speckle suppression techniques based on the wavelet transform performs the best, followed by the modified K-nearest neighbours and the Lee's local statistic filters. Depending on the nature of the land cover types being classified, texture features derived from second- and third-order histogram performed the best, followed by first-order statistics and features derived using the grey-level difference vector method. Among all classifiers considered, the MLP and the RBF neural networks performed the best, achieving up to 94% overall accuracy for the E-SAR image of Douala, for example.

Land-cover classification methods for multi-year AVHRR data

Advanced Very High Resolution Radiometer (AVHRR) data have been extensively used for global land-cover classification, but few studies have taken direct and full advantage of the multi-year properties of AVHRR data. This study focused on generating effective classification features from multi-year AVHRR data to improve classification accuracy. Three types of features were derived from 12-year monthly composite normalized difference vegetation index (NDVI) and channel 4 brightness temperature from the NOAA/NASA Pathfinder AVHRR Land data for land-cover classification. The first is based on the shape of the annual average NDVI or brightness-temperature profile, which was then approximated by a Fourier series. The coefficients estimated by the weighted least-squares method were used for classification. The second and third features were based on the raw periodogram of the time series and the auto-regressive modelling. A global land-cover training database created from Landsat Thematic Mapper and Multi-spectr...

ミクセルの存在とトレーニングデータの代表性を考慮した時系列低空間分解能画像を用いた土地被覆分類手法

This paper describes a land cover classification method, which uses time-series and low spatial resolution satellite images, and which can discriminate the feature space with high reliability for classification from the other. The method decomposes a whole of image histogram into each component, feature distributions of pure classes and mixel class. The decomposition result gives the likelihood function for a specific land cover class, then the feature space is discriminated based on the likelihood function. Validation demonstrates that the presented method is effective to improve the representativeness of training data sets.

Land-cover classification methods for multi-year AVHRR data

Advanced Very High Resolution Radiometer (AVHRR) data have been extensively used for global land-cover classification, but few studies have taken direct and full advantage of the multi-year properties of AVHRR data. This study focused on generating effective classification features from multi-year AVHRR data to improve classification accuracy. Three types of features were derived from 12-year monthly composite normalized difference vegetation index (NDVI) and channel 4 brightness temperature from the NOAA/NASA Pathfinder AVHRR Land data for land-cover classification. The first is based on the shape of the annual average NDVI or brightness-temperature profile, which was then approximated by a Fourier series. The coefficients estimated by the weighted least-squares method were used for classification. The second and third features were based on the raw periodogram of the time series and the auto-regressive modelling. A global land-cover training database created from Landsat Thematic Mapper and Multi-spectral Scanner imagery was used for training and validation. Both quadrature discriminate analysis (QDA) and linear discriminate analysis (LDA) were explored for classification, and results indicate that QDA performs much better than LDA. The first feature, based on the mean annual shape, produced much better results than the other two features. It was also found that NDVI signals worked better than brightness-temperature signals. That is probably because topof-atmosphere signals were used, and atmospheric contaminations significantly disturb the thermal signals and correlation structures of different cover types. Further validations are needed.