Land Cover Classification Information Research Articles

Multimodal self-supervised learning for remote sensing data land cover classification

Deep learning has revolutionized the remote sensing image processing techniques over the past few years. Nevertheless, annotating high-quality samples is difficult and time-consuming, which limits the performance of deep neural networks because of insufficient supervision information. Aiming to solve this contradiction, we investigate the multimodal self-supervised learning (MultiSSL) paradigm for pre-training and classification of remote sensing image. Specifically, the proposed self-supervised feature learning model consists of asymmetric encoder–decoder structure, in which deep unified encoder learns high-level key information characterizing multimodal remote sensing data and task-specific lightweight decoders are developed to reconstruct original data. To further enhance feature extraction capability, the cross-attention layers are utilized to exchange information contained in heterogeneous characteristics, thus learning more complementary information from multimodal remote sensing data. In fine-tuning stage, the pre-trained encoder and cross-attention layer serve as feature extractor, and leaned characteristics are combined with corresponding spectral information for land cover classification through a lightweight classifier. The self-supervised pre-training model can learn high-level key features from unlabeled samples, thereby utilizing the feature extraction capability of deep neural networks while reducing their dependence on annotated samples. Compared with existing classification paradigms, the proposed multimodal self-supervised pre-training and fine-tuning scheme achieves superior performance for remote sensing image land cover classification.

Unsupervised Domain Adaptation with Contrastive Learning-Based Discriminative Feature Augmentation for RS Image Classification

High- and very high-resolution (HR, VHR) remote sensing (RS) images can provide comprehensive and intricate spatial information for land cover classification, which is particularly crucial when analyzing complex built-up environments. However, the application of HR and VHR images to large-scale and detailed land cover mapping is always constrained by the intricacy of land cover classification models, the exorbitant cost of collecting training samples, and geographical changes or acquisition conditions. To overcome this limitation, we propose an unsupervised domain adaptation (UDA) with contrastive learning-based discriminative feature augmentation (CLDFA) for RS image classification. In detail, our method first utilizes contrastive learning (CL) through a memory bank in order to memorize sample features and improve model performance, where the approach employs an end-to-end Siamese network and incorporates dynamic pseudo-label assignment and class-balancing strategies for adaptive domain joint learning. By transferring classification models trained on a source domain (SD) to an unlabeled target domain (TD), our proposed UDA method enables large-scale land cover mapping. We conducted experiments using a massive five billion-pixels dataset as the SD and tested the HR and VHR RS images of five typical Chinese cities as the TD and applied the method on the completely unlabeled world view 3 (WV3) image of Urumqi city. The experimental results demonstrate that our method excels in large-scale HR and VHR RS image classification tasks, highlighting the advantages of semantic segmentation based on end-to-end deep convolutional neural networks (DCNNs).

Application of Sentinel-2A Images for Land Cover Classification Using NDVI in Jember Regency

The advancement of remote sensing technology has led to the development of sophisticated image processing methods that yield highly accurate land cover classification information, minimizing misinterpretations. The Normalized Difference Vegetation Index (NDVI) is a widely utilized method in remote sensing for measuring green vegetation. A significant portion of the Jember Regency area is covered by vegetation. This study aimed to identify various land cover types in the Jember Regency area, quantify the area for each classification, and establish the NDVI value ranges for each type of cover. Sentinel-2 was employed as the primary data source, and the NDVI method was utilized for land cover classification in the Jember Regency. The region exhibited diverse land cover types. Data from Sentinel-2A captured in June and October 2019 were chosen due to their accessibility, open-source nature, and adequate spectral, spatial, and temporal resolution. The classification in this study encompassed five classes: water bodies, settlements, dry fields, irrigated paddy fields, and forests. Error analysis was conducted using a confusion matrix with the Overall and Kappa algorithms. The accuracy results for June indicated a Kappa Accuracy of 37.7% and Overall Accuracy of 54.5%. In October, the Kappa Accuracy increased to 39.9%, and the Overall Accuracy reached 56.5%. In conclusion, the NDVI method did not meet the criteria for accurately interpreting land cover classification.

Rapid Land Cover Classification Using a 36-Year Time Series of Multi-Source Remote Sensing Data

Long time series land cover classification information is the basis for scientific research on urban sprawls, vegetation change, and the carbon cycle. The rapid development of cloud computing platforms such as the Google Earth Engine (GEE) and access to multi-source satellite imagery from Landsat and Sentinel-2 enables the application of machine learning algorithms for image classification. Here, we used the random forest algorithm to quickly achieve a time series land cover classification at different scales based on the fixed land classification sample points selected from images acquired in 2022, and the year-by-year spectral differences of the sample points. The classification accuracy was enhanced by using multi-source remote sensing data, such as synthetic aperture radar (SAR) and digital elevation model (DEM) data. The results showed that: (i) the maximum difference (threshold) of the sample points without land class change, determined by counting the sample points of each band of the Landsat time series from 1986 to 2022, was 0.25; (ii) the kappa coefficient and observed accuracy of the same sensor from Landsat 8 are higher than the results of the TM and ETM+ sensor data from 2013 to 2022; and (iii) the addition of a mining land cover type increases the kappa coefficient and overall accuracy mean values of the Sentinel 2 image classification for a complex mining and forest area. Among the land classifications via multi-source remote sensing, the combined variables of Spectral band + Index + Terrain + SAR result in the highest accuracy, but the overall improvement is limited. The method proposed is applicable to remotely sensed images at different scales and the use of sensors under complex terrain conditions. The use of the GEE cloud computing platform enabled the rapid analysis of remotely sensed data to produce land cover maps with high accuracy and a long time series.

Enabling country-scale land cover mapping with meter-resolution satellite imagery

High-resolution satellite images can provide abundant, detailed spatial information for land cover classification, which is particularly important for studying the complicated built environment. However, due to the complex land cover patterns, the costly training sample collections, and the severe distribution shifts of satellite imageries caused by, e.g., geographical differences or acquisition conditions, few studies have applied high-resolution images to land cover mapping in detailed categories at large scale. To fill this gap, we present a large-scale land cover dataset, Five-Billion-Pixels. It contains more than 5 billion labeled pixels of 150 high-resolution Gaofen-2 (4 m) satellite images, annotated in a 24-category system covering artificial-constructed, agricultural, and natural classes. In addition, we propose a deep-learning-based unsupervised domain adaptation approach that can transfer classification models trained on labeled dataset (referred to as the source domain) to unlabeled data (referred to as the target domain) for large-scale land cover mapping. Specifically, we introduce an end-to-end Siamese network employing dynamic pseudo-label assignment and class balancing strategy to perform adaptive domain joint learning. To validate the generalizability of our dataset and the proposed approach across different sensors and different geographical regions, we carry out land cover mapping on five megacities in China and six cities in other five Asian countries severally using: PlanetScope (3 m), Gaofen-1 (8 m), and Sentinel-2 (10 m) satellite images. Over a total study area of 60,000 km2, the experiments show promising results even though the input images are entirely unlabeled. The proposed approach, trained with the Five-Billion-Pixels dataset, enables high-quality and detailed land cover mapping across the whole country of China and some other Asian countries at meter-resolution.

Fusing Ultra-Hyperspectral and High Spatial Resolution Information for Land Cover Classification Based on AISAIBIS Sensor and Phase Camera

Hyperspectral imaging technology are widely used in vegetation, agriculture, and other fields, especially in land cover classification of complex scenes. Higher spectral resolution has become the focus of the development of hyperspectral imaging technology for classification. The advent of airborne AISAIBIS sensor reaches 0.11 nm ultrahyperspectral resolution. The ultrahyperspectral imagery shows great advantages in classification with its increasing spectral resolution. But its spatial resolution is limited because of the imaging mechanism, which brings great difficulties to the accurate extract of fine and regular objects. Therefore, we proposed an optimal fusion and classification strategy based on the complementary advantage information of ultrahyperspectral and high spatial resolution image. The fusion feasibility and effectiveness were verified by various fusion methods. And a quality evaluation system was developed to assess the quality of fusion results. Besides, a multiresolution segmentation optimization and classification evaluation scheme was proposed to comparatively analyze the effect of optimal fusion result on improving classification accuracy. Results show that the classification accuracy of the optimal fused image reaches 88.10%, and 7.11%–19.03% higher than that of original images. It fully validates the effectiveness of the strategy proposed in this article.

UAV-Based Multi-Sensor Data Fusion for Urban Land Cover Mapping Using a Deep Convolutional Neural Network

Accurate and up-to-date land cover classification information is essential for many applications, such as land-use change detection, global environmental change, and forest management, among others. Unoccupied aerial systems (UASs) provide the advantage of flexible and rapid data acquisition at low cost compared to conventional platforms, such as satellite and airborne systems. UASs are often equipped with high spatial resolution cameras and/or light detection and ranging (LiDAR). However, the high spatial resolution imagery has a high information content, which makes land cover classification quite challenging. Recently, deep convolutional neural networks (DCNNs) have been effectively applied to remote sensing applications, which overcome the drawback of traditional techniques. In this research, a low-cost UAV-based multi-sensor data fusion model was developed for land cover classification based on a DCNN. For the purpose of this research, two datasets were collected at two different urban locations using two different UASs. A DCNN model, based on U-net with Resnet101 as a backbone, was used to train and test the fused image/LiDAR data. The maximum likelihood and support vector machine techniques were used as a reference for classifier comparison. It was shown that the proposed DCNN approach improved the overall accuracy of land cover classification for the first dataset by 15% compared to the reference classifiers. In addition, the overall accuracy of land cover classification improved by 7%, and the precision, recall, and F-measure improved by 18% when the fused image/LiDAR data were used compared to the images only. The trained DCNN model was also tested on the second dataset, and the obtained results were largely similar to those of the first dataset.

Classification Airbrone LiDAR Point cloud Data

Land cover classification information plays a very important role in various applications. Airborne Light detection and Ranging (LiDAR) data is widely used in remote sensing application for the classification of land cover. The present study presents a Spatial classification method using Terrasoild macros . The data used in this study are a LiDAR point cloud data with the wavelength of green:532nm, near infrared:1064nm and mid-infrared-1550nm and High Resolution RGB data. The classification is carried in TERRASCAN Module with twelve land cover classes. The classification accuracies were assessed using high resolution RGB data. From the results it is concluded that the LiDAR data classification with overall accuracy and kappa coefficient 85.2% and 0.7562.

INTER-ANNUAL DIFFERENCE OF LAND SURFACE PHENOLOGY (LSP) TRANSACTION DATES USING TIME SERIES LANDSAT IMAGES

Abstract. Land surface phenology (LSP) is a kind of vital information for land cover classification and vegetation growth monitoring. Time series Landsat images, with the advantages of long observations and high spatial resolution, have been widely used in LSP identification. However, LSP transaction dates, such as start of season (SOS) and end of season (EOS), are highly influenced by the coarse temporal resolution. In this study, we compare the inter-annual difference of LSP SOS from 5 years interval, 10 years interval and all years interval Landsat images, and improve the SOS estimated model by considering the accumulated growing degree-days (AGDD) of soil temperature and soil moisture. Results indicate that LSP SOS can serve as a good proxy for reflecting ground vegetation phenology, especially using 5 years interval Landsat images. Soil temperature and soil moisture have certain influence on SOS estimation, and the R-squared value reached 0.9 after model adjustment. This study can provide guidance for estimating suitable inter-annual LSP transaction dates under different sceneries in the future.

Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery

This study compares some different types of spectral domain transformations for convolutional neural network (CNN)-based land cover classification. A novel approach was proposed, which transforms one-dimensional (1-D) spectral vectors into two-dimensional (2-D) features: Polygon graph images (CNN-Polygon) and 2-D matrices (CNN-Matrix). The motivations of this study are that (1) the shape of the converted 2-D images is more intuitive for human eyes to interpret when compared to 1-D spectral input; and (2) CNNs are highly specialized and may be able to similarly utilize this information for land cover classification. Four seasonal Landsat 8 images over three study areas—Lake Tapps, Washington, Concord, New Hampshire, USA, and Gwangju, Korea—were used to evaluate the proposed approach for nine land cover classes compared to several other methods: Random forest (RF), support vector machine (SVM), 1-D CNN, and patch-based CNN. Oversampling and undersampling approaches were conducted to examine the effect of the sample size on the model performance. The CNN-Polygon had better performance than the other methods, with overall accuracies of about 93%–95 % for both Concord and Lake Tapps and 80%–84% for Gwangju. The CNN-Polygon particularly performed well when the training sample size was small, less than 200 per class, while the CNN-Matrix resulted in similar or higher performance as sample sizes became larger. The contributing input variables to the models were carefully analyzed through sensitivity analysis based on occlusion maps and accuracy decreases. Our result showed that a more visually intuitive representation of input features for CNN-based classification models yielded higher performance, especially when the training sample size was small. This implies that the proposed graph-based CNNs would be useful for land cover classification where reference data are limited.

On the Effects of InSAR Temporal Decorrelation and Its Implications for Land Cover Classification: The Case of the Ocean-Reclaimed Lands of the Shanghai Megacity.

In this work, we focused on the ocean-reclaimed lands of the Shanghai coastal region and we evidenced how, over these areas, the interferometric synthetic aperture radar (InSAR) coherence maps exhibit peculiar behavior. In particular, by analyzing a sequence of Sentinel-1 SAR InSAR coherence maps, we found a significant coherence loss over time in correspondence to the ocean-reclaimed platforms that are substantially different from the coherence loss experienced in naturally-formed regions with the same type of land cover. We have verified whether this is due to the engineering geological conditions or the soil consolidation subsidence in ocean-reclaimed region. In this work, we combine the information coming from InSAR coherence maps and the retrieved temporal decorrelation model with that obtained by using optical Sentinel-2 data, and we performed land cover classification analyses in the zone of the Pudong International Airport. To estimate the accuracy of utilizing InSAR coherence information for land cover classification, in particular, we have analyzed what causes the difference of the InSAR coherence loss with the same type of land cover. The presented results show that the coherence models can be useful to distinguish roads and buildings, thus enhancing the accuracy of land cover classification compared with that allowable by using only Sentinel-2 data. In particular, the accuracy of classification increases from 75% to 86%.

Land cover classification using landsat-8 optical data and supervised classifiers

Land use and land cover information of an area has got importance in various aspects mainly because of various development activities that are taking place in every part of the world. Various satellite sensors are providing the required data collected by remote sensing techniques in the form of images using which the land use land cover information can be analyzed. Constistency of Landsat satellite is illustrated with two time periods such as Operational Land Imager (OLI) of 2013 and consecutive 2014 procured by earth explorer with quantified changes for the same period in visakhapatnam of hudhud cyclone. Since this city is consisting of mainly urban, vegetation, few water bodies, some area of agriculture and barren,five classes have been chosen from the study area. The results indicate that due to the hudhud event some changes took place. vegetation and built-up land have been increased by An increase of 19.1% (6.3 km2) and 11% (5.36 km2) has been observed in the case of vegetation and built up area where as a decrease of 1.2% (4.06 km2), 6.1% (1.70 km2) and 1.2% (0.72 km2) has been observed in the case of agriculture, barren land, and water body respectively. With the help of available satellite imagery belonging to the same area and of different time periods along with the change detection techniques landscape dynamics have been analyzed. Using various classification algorithms along with the data available from the satellite sensor the land use and land cover classification information of the study area has been obtained. The maximum likelihood algorithm provided better results compared to other classification techniques and the accuracy achieved with this algorithm is 99.930% (overall accuracy) and 0.999 (Kappa coefficient).

Fusion of texture and wavelet features of PALSAR image using LDA and PCA for land cover classification

ABSTRACTLand cover classification is yet a challenging task due to complex landscapes, less a priori knowledge and complicated data sets. Researchers are utilising scattering-based model, polarimetric-based model and empirical model for classification purpose on SAR (Synthetic Aperture Radar) data, but achieving good classification accuracy requires more consideration. Most of the work reported has been performed by either utilising only single polarisation channel or on the composite image, due to which the significance of each channel remains unidentified. Previously, SAR feature-based land cover characterisation has shown favourable results, by examining the use of one or two forms of features and not much attempt has been dedicated to the simultaneous incorporation of multiple types of texture and wavelet features in a manner that preserves all the significant segregation information for land cover classification. Therefore, in this paper, different statistical, textural and wavelet features were analysed on three individual polarimetric channels (HH, HV and VV). For enhancing the discrimination capability, principal component analysis and linear discriminant analysis have been employed and the classification information attained from the three polarimetric channels was fused for accounting the implication of each of them. Proposed methodology is screening decent classification accuracy of 80.83% with kappa coefficient 0.7289, with less complexity and easy implementation.

Volume and Carbon Estimates for the Forest Area of the Amhara Region in Northwestern Ethiopia

Sustainable forest management requires a continuous assessment of the forest conditions covering the species distribution, standing tree volume as well as volume increment rates. Forest inventories are designed to record this information. They, in combination with ecosystem models, are the conceptual framework for sustainable forest management. While such management systems are common in many countries, no forest inventory system and/or modeling tools for deriving forest growth information are available in Ethiopia. This study assesses, for the first time, timber volume, carbon, and Net Primary Production (NPP) of forested areas in the Amhara region of northwestern Ethiopia by combining (i) terrestrial inventory data, and (ii) land cover classification information. The inventory data were collected from five sites across the Amhara region (Ambober, Gelawdiwos, Katassi, Mahiberesilasse and Taragedam) covering three forest types: (i) forests, (ii) shrublands (exclosures) and (ii) woodlands. The data were recorded on 198 sample plots and cover diameter at breast height, tree height, and increment information. In order to extrapolate the local terrestrial inventory data to the whole Amhara region, a digital land cover map from the Amhara’s Bureau of Agriculture was simplified into (i) forest, (ii) shrubland, and (iii) woodland. In addition, the forest area is further stratified in five elevation classes. Our results suggest that the forest area in the Amhara region covers 2% of the total land area with an average volume stock of 65.7 m3·ha−1; the shrubland covers 27% and a volume stock of 3.7 m3·ha−1; and the woodland covers 6% and an average volume stock of 27.6 m3·ha−1. The corresponding annual volume increment rates are 3.0 m3·ha−1, for the forest area; 1.0 m3·ha−1, for the shrubland; and 1.2 m3·ha−1, for the woodland. The estimated current total volume stock in the Amhara region is 59 million m3.

Fusion of Airborne Discrete-Return LiDAR and Hyperspectral Data for Land Cover Classification

Accurate land cover classification information is a critical variable for many applications. This study presents a method to classify land cover using the fusion data of airborne discrete return LiDAR (Light Detection and Ranging) and CASI (Compact Airborne Spectrographic Imager) hyperspectral data. Four LiDAR-derived images (DTM, DSM, nDSM, and intensity) and CASI data (48 bands) with 1 m spatial resolution were spatially resampled to 2, 4, 8, 10, 20 and 30 m resolutions using the nearest neighbor resampling method. These data were thereafter fused using the layer stacking and principal components analysis (PCA) methods. Land cover was classified by commonly used supervised classifications in remote sensing images, i.e., the support vector machine (SVM) and maximum likelihood (MLC) classifiers. Each classifier was applied to four types of datasets (at seven different spatial resolutions): (1) the layer stacking fusion data; (2) the PCA fusion data; (3) the LiDAR data alone; and (4) the CASI data alone. In this study, the land cover category was classified into seven classes, i.e., buildings, road, water bodies, forests, grassland, cropland and barren land. A total of 56 classification results were produced, and the classification accuracies were assessed and compared. The results show that the classification accuracies produced from two fused datasets were higher than that of the single LiDAR and CASI data at all seven spatial resolutions. Moreover, we find that the layer stacking method produced higher overall classification accuracies than the PCA fusion method using both the SVM and MLC classifiers. The highest classification accuracy obtained (OA = 97.8%, kappa = 0.964) using the SVM classifier on the layer stacking fusion data at 1 m spatial resolution. Compared with the best classification results of the CASI and LiDAR data alone, the overall classification accuracies improved by 9.1% and 19.6%, respectively. Our findings also demonstrated that the SVM classifier generally performed better than the MLC when classifying multisource data; however, none of the classifiers consistently produced higher accuracies at all spatial resolutions.

Large-scale land cover mapping with the integration of multi-source information based on the Dempster–Shafer theory

Land cover type is a crucial parameter that is required for various land surface models that simulate water and carbon cycles, ecosystem dynamics, and climate change. Many land use/land cover maps used in recent years have been derived from field investigations and remote-sensing observations. However, no land cover map that is derived from a single source (such as satellite observation) properly meets the needs of land surface simulation in China. This article presents a decision-fuse method to produce a higher-accuracy land cover map by combining multi-source local data based on the Dempster–Shafer (D–S) evidence theory. A practical evidence generation scheme was used to integrate multi-source land cover classification information. The basic probability values of the input data were obtained from literature reviews and expert knowledge. A Multi-source Integrated Chinese Land Cover (MICLCover) map was generated by combining multi-source land cover/land use classification maps including a 1:1,000,000 vegetation map, a 1:100,000 land use map for the year 2000, a 1:1,000,000 swamp-wetland map, a glacier map, and a Moderate-Resolution Imaging Spectroradiometer land cover map for China in 2001 (MODIS2001). The merit of this new map is that it uses a common classification system (the International Geosphere-Biosphere Programme (IGBP) land cover classification system), and it has a unified 1 km resolution. The accuracy of the new map was validated by a hybrid procedure. The validation results show great improvement in accuracy for the MICLCover map. The local-scale visual comparison validations for three regions show that the MICLCover map provides more spatial details on land cover at the local scale compared with other popular land cover products. The improvement in accuracy is true for all classes but particularly for cropland, urban, glacier, wetland, and water body classes. Validation by comparison with the China Forestry Scientific Data Center (CFSDC)–Forest Inventory Data (FID) data shows that overall forest accuracies in five provinces increased to between 42.19% and 88.65% for our MICLCover map, while those of the MODIS2001 map increased between 27.77% and 77.89%. The validation all over China shows that the overall accuracy of the MICLCover map is 71%, which is higher than the accuracies of other land cover maps. This map therefore can be used as an important input for land surface models of China. It has the potential to improve the modeling accuracy of land surface processes as well as to support other aspects of scientific land surface investigations in China.

Object‐oriented and textural image classification of the Siberia GBFM radar mosaic combined with MERIS imagery for continental scale land cover mapping

Boreal forests and wetlands play an important role in the climate system, in particular through biosphere–atmosphere flux exchanges. They are an important pool of carbon and their role as sink or source of greenhouse gases is not fully understood. Accurate mapping of the vegetation of Siberia can therefore contribute to a better understanding of these processes at regional scale and of their effects on the climate through regional biosphere modeling. The potential of the combination of radar data with medium‐resolution optical data to obtain regional‐scale land cover mapping is investigated using multi‐spectral imagery from the MERIS sensor at 300 m resolution and a high resolution radar mosaic (pixel spacing of 100 m) covering Western and Eastern Siberia compiled in the framework of the Global Boreal Forest Mapping project. For this purpose, capabilities of oriented‐object image analysis associated to wavelet multi‐resolution techniques are investigated. Results show that wavelet multi‐resolution textures bring relevant additional information for land cover classification. Suggestions are made for the implementation of an object‐based wavelet multi‐resolution texture estimator.

Multisource Data Fusion for Landslide Classification Using Generalized Positive Boolean Functions

In this paper, a novel technique is proposed for a supervised classification of multisource images for the purpose of landslide hazard assessment. The method, known as the generalized positive Boolean function (GPBF), is developed for land cover classification based on the fusion of remotely sensed images of the same scene collected from multiple sources. It presents a framework for data fusion of multisource remotely sensed images, which consists of two approaches, referred to as the band generation process (BGP) and the positive Boolean function (PBF) classifier. The PBF classifier developed from a stack filter has been successfully applied in hyperspectral image classification. For the PBF to be effective for multispectral images, a multiple adaptation BGP is introduced to create a new set of additional bands especially accommodated to landslide classes. These bands include nonlinear normalized difference vegetation index data and morphological information in the form of digital elevation model (DEM)-derived slope values that originate from multiple sources. The performance of the proposed method is evaluated by fusing Systeme Pour l'Observation de la Terre images and DEM information for land cover classification during the post 921 Earthquake period in Taiwan. Experimental results demonstrate the proposed GPBF multiclassification approach is suitable for land cover classification in Earth remote sensing and improves the precision of image classification compared to conventional classifiers

Land cover classification using MODIS–ASTER airborne simulator (MASTER) data and NDVI: a case study of the Kochang area, Korea

New sensors and new technologies for data processing provide more capabilities to map and extract information about land cover and land use. MODIS–ASTER airborne simulator (MASTER) data acquired during the PACRIM II mission were used for land cover classification in the Kochang area, Korea. Twenty-two bands covering the visible, near-infrared, and shortwave infrared wavelengths were used in the study. A minimum noise fraction (MNF) transform was used for feature extraction to select an optimum subset of data in terms of classification accuracy. The first six MNF images obtained the most accurate classification result among all MNF data combinations. Classification using five normalized difference vegetation index (NDVI) images from the visible red band and five different near-infrared bands achieved a relatively good result (overall accuracy of 78.6%) and showed additional discriminative information for land cover classification. The integration of spectral information and NDVI data using a hierarchical classification method substantially improved the classification accuracy compared to classification with spectral data alone (from 84.35% to 90.62%). The knowledge from visual inspection of classification results and analysis of confusion matrices was used to build rules used in the hierarchical classification.