Very High Resolution Images Research Articles

BUILDING CLASSIFICATION OF VHR AIRBORNE STEREO IMAGES USING FULLY CONVOLUTIONAL NETWORKS AND FREE TRAINING SAMPLES

Abstract. Semantic segmentation, especially for buildings, from the very high resolution (VHR) airborne images is an important task in urban mapping applications. Nowadays, the deep learning has significantly improved and applied in computer vision applications. Fully Convolutional Networks (FCN) is one of the tops voted method due to their good performance and high computational efficiency. However, the state-of-art results of deep nets depend on the training on large-scale benchmark datasets. Unfortunately, the benchmarks of VHR images are limited and have less generalization capability to another area of interest. As existing high precision base maps are easily available and objects are not changed dramatically in an urban area, the map information can be used to label images for training samples. Apart from object changes between maps and images due to time differences, the maps often cannot perfectly match with images. In this study, the main mislabeling sources are considered and addressed by utilizing stereo images, such as relief displacement, different representation between the base map and the image, and occlusion areas in the image. These free training samples are then fed to a pre-trained FCN. To find the better result, we applied fine-tuning with different learning rates and freezing different layers. We further improved the results by introducing atrous convolution. By using free training samples, we achieve a promising building classification with 85.6 % overall accuracy and 83.77 % F1 score, while the result from ISPRS benchmark by using manual labels has 92.02 % overall accuracy and 84.06 % F1 score, due to the building complexities in our study area.

Hierarchical spatial features learning with deep CNNs for very high-resolution remote sensing image classification

ABSTRACTRecently, very high resolution (VHR) remote sensing (RS) images have become widely available. While a VHR RS image provides more detailed information of the imaged scene, inter-class variation is enhanced as well, which makes it a challenging task to accurately classify the image at pixel level. It has been found that spatial features of a VHR image derived from the spectral measurements can improve the accuracy of image classification. Recent studies have shown that deep convolutional neural networks (CNNs) are powerful to extract high-level spatial features for object recognition and image retrieval. However, few CNNs have been developed for the pixel-wise classification of RS image. Moreover, in order to obtain large-scale spatial features, CNNs can be unmanageably complicated. In this paper, we propose a novel deep CNNs based method for VHR image classification. Firstly, pixel level CNNs are applied to multiscale images generated from the original VHR RS image using the Gaussian pyramid method. The hierarchical multiscale spatial features are then resampled to the size of the raw image. Finally, feature fusion is conducted with the original spectral bands for spectral-spatial based classification, where a support vector machine (SVM) classifier can be employed. Experiments were conducted on the image of VHR aerial image in Vaihingen, Germany. Compared to using spectral features only or with extended morphological profiles (EMPs) for image classification, the proposed method performs better, and the best improvement of classification accuracy is as high as 13%.

Refining Land Cover Classification Maps Based on Dual-Adaptive Majority Voting Strategy for Very High Resolution Remote Sensing Images

Land cover classification that uses very high resolution (VHR) remote sensing images is a topic of considerable interest. Although many classification methods have been developed, the accuracy and usability of classification systems can still be improved. In this paper, a novel post-processing approach based on a dual-adaptive majority voting strategy (D-AMVS) is proposed to improve the performance of initial classification maps. D-AMVS defines a strategy for refining each label of a classified map that is obtained by different classification methods from the same original image, and fusing the different refined classification maps to generate a final classification result. The proposed D-AMVS contains three main blocks. (1) An adaptive region is generated by gradually extending the region around a central pixel based on two predefined parameters (T1 and T2) to utilize the spatial feature of ground targets in a VHR image. (2) For each classified map, the label of the central pixel is refined according to the majority voting rule within the adaptive region. This is defined as adaptive majority voting. Each initial classified map is refined in this manner pixel by pixel. (3) Finally, the refined classified maps are used to generate a final classification map, and the label of the central pixel in the final classification map is determined by applying AMV again. Each entire classified map is scanned and refined pixel by pixel based on the proposed D-AMVS. The accuracies of the proposed D-AMVS approach are investigated with two remote sensing images with high spatial resolutions of 1.0 m and 1.3 m. Compared with the classical majority voting method and a relatively new post-processing method called the general post-classification framework, the proposed D-AMVS can achieve a land cover classification map with less noise and higher classification accuracies.

Modelling relational contexts in GEOBIA framework for improving urban land-cover mapping

ABSTRACTIn human cognition, both visual features (i.e., spectrum, geometry and texture) and relational contexts (i.e. spatial relations) are used to interpret very-high-resolution (VHR) images. However, most existing classification methods only consider visual features, thus classification performances are susceptible to the confusion of visual features and the complexity of geographic objects in VHR images. On the contrary, relational contexts between geographic objects are some kinds of spatial knowledge, thus they can help to correct initial classification errors in a classification post-processing. This study presents the models for formalizing relational contexts, including relative relations (like alongness, betweeness, among, and surrounding), direction relation (azimuth) and their combination. The formalized relational contexts were further used to define locally contextual regions to identify those objects that should be reclassified in a post-classification process and to improve the results of an initial classification. The experimental results demonstrate that the relational contexts can significantly improve the accuracies of buildings, water, trees, roads, other surfaces and shadows. The relational contexts as well as their combinations can be regarded as a contribution to post-processing classification techniques in GEOBIA framework, and help to recognize image objects that cannot be distinguished in an initial classification.

Urban Change Detection Based on Dempster–Shafer Theory for Multitemporal Very High-Resolution Imagery

Fusing multiple change detection results has great potentials in dealing with the spectral variability in multitemporal very high-resolution (VHR) remote sensing images. However, it is difficult to solve the problem of uncertainty, which mainly includes the inaccuracy of each candidate change map and the conflicts between different results. Dempster–Shafer theory (D–S) is an effective method to model uncertainties and combine multiple evidences. Therefore, in this paper, we proposed an urban change detection method for VHR images by fusing multiple change detection methods with D–S evidence theory. Change vector analysis (CVA), iteratively reweighted multivariate alteration detection (IRMAD), and iterative slow feature analysis (ISFA) were utilized to obtain the candidate change maps. The final change detection result is generated by fusing the three evidences with D–S evidence theory and a segmentation object map. The experiment indicates that the proposed method can obtain the best performance in detection rate, false alarm rate, and comprehensive indicators.

3D BUILDING CHANGE DETECTION BETWEEN CURRENT VHR IMAGES AND PAST LIDAR DATA

Abstract. Change detection is an essential step to locate the area where an old model should be updated. With high density and accuracy, LiDAR data is often used to create a 3D city model. However, updating LiDAR data at state or nation level often takes years. Very high resolution (VHR) images with high updating rate is therefore an option for change detection. This paper provides a novel and efficient approach to derive pixel-based building change detection between past LiDAR and new VHR images. The proposed approach aims notably at reducing false alarms of changes near edges. For this purpose, LiDAR data is used to supervise the process of finding stereo pairs and derive the changes directly. This paper proposes to derive three possible heights (so three DSMs) by exploiting planar segments from LiDAR data. Near edges, the up to three possible heights are transformed into discrete disparities. A optimal disparity is selected from a reasonable and computational efficient range centered on them. If the optimal disparity is selected, but still the stereo pair found is wrong, a change has been found. A Markov random field (MRF) with built-in edge awareness from images is designed to find optimal disparity. By segmenting the pixels into plane and edge segments, the global optimization problem is split into many local ones which makes the optimization very efficient. Using an optimization and a consecutive occlusion consistency check, the changes are derived from stereo pairs having high color difference. The algorithm is tested to find changes in an urban areas in the city of Amersfoort, the Netherlands. The two different test cases show that the algorithm is indeed efficient. The optimized disparity images have sharp edges along those of images and false alarms of changes near or on edges and occlusions are largely reduced.

A GUIDED REGISTRATION STRATEGY EMPLOYING VIRTUAL PLANES TO OVERCOME NON-STANDARD GEOMETRIES – USING THE EXAMPLE OF MOBILE MAPPING AND AERIAL OBLIQUE IMAGERY

Abstract. Mobile mapping (MM) is an intriguing as well as emerging platform and technology for geo-data acquisition. In typical areas of interest for MM campaigns, such as urban areas, unwanted GNSS multipath, non-line-of-sight effects, and IMU drifts may lead to deteriorated position fixes. In this work, we are proposing a novel technique to register MM and aerial oblique imagery. As aerial platforms are not affected by GNSS occlusions and are able to collect very-high resolution images, a co-registration of the data sets enables a) an independent verification of the platform’s accuracy and b) an adjustment of the MM data’s pose. Both data sets depict the scene from an entirely different perspective, which complicates the matching problem. Our approach is based on the assumption that common visible entities in both images are available, e.g. façade surfaces. By determining planes coinciding with these visible entities in object space, variances can be overcome. As the orientation of the data sets is known – MM data has an unknown accuracy – derived planes are employed to support a visibility hypothesis while storing image information for image registration in object space. This enables constraining search space to support the actual registration. Although the inhomogeneity of the data sets poses a challenge to a successful registration, we can show that our stepwise strategy of finding common visible entities, exploiting them to increase the resemblance of the data sets, and utilising accurate registration methods renders this matching scenario possible. In this paper, the algorithm is explained in detail, experimental results of significant steps will be shown, and possible extensions are discussed.

Relative radiometric normalization of bitemporal very high-resolution satellite images for flood change detection

Change-detection analysis using bitemporal satellite imagery is a reliable method for providing and assessing information about flood-induced changes over a wide area in a timely and cost-effective manner. Accurate radiometric normalization between bitemporal imagery is a critical component in the application of change-detection techniques to flood mapping because the accuracy of the change detection is directly affected by the quality of radiometric normalization. A methodology based on multivariate alteration detection (MAD) is introduced as an approach that enables reliable radiometric normalization of bitemporal very high-resolution (VHR) images for detecting flood-induced changes. The method uses a weighting function to adaptively identify weights based on open water features, which are estimated by the normalized difference water index, in the computation of the covariance matrices of the MAD transform. To quantitatively evaluate and test the performance of the proposed method, a comparison is made between it and the iteratively reweighted (IR)-MAD method based on statistical tests and the accuracy of flood change detection. Change vector analysis- and MAD-based change-detection methods were used for the comparison of the proposed and IR-MAD methods. Experimental results on KOMPSAT-2 bitemporal VHR images prove that the proposed method produced better results than the IR-MAD method in the statistical tests and also increased the overall accuracy of flood change detection by 1.8% and 12.6% for the two study sites.

PURIFICATION OF TRAINING SAMPLES BASED ON SPECTRAL FEATURE AND SUPERPIXEL SEGMENTATION

Abstract. Remote sensing image classification is an effective way to extract information from large volumes of high-spatial resolution remote sensing images. Generally, supervised image classification relies on abundant and high-precision training data, which is often manually interpreted by human experts to provide ground truth for training and evaluating the performance of the classifier. Remote sensing enterprises accumulated lots of manually interpreted products from early lower-spatial resolution remote sensing images by executing their routine research and business programs. However, these manually interpreted products may not match the very high resolution (VHR) image properly because of different dates or spatial resolution of both data, thus, hindering suitability of manually interpreted products in training classification models, or small coverage area of these manually interpreted products. We also face similar problems in our laboratory in 21st Century Aerospace Technology Co. Ltd (short for 21AT). In this work, we propose a method to purify the interpreted product to match newly available VHRI data and provide the best training data for supervised image classifiers in VHR image classification. And results indicate that our proposed method can efficiently purify the input data for future machine learning use.

An Approach for Unsupervised Change Detection in Multitemporal VHR Images Acquired by Different Multispectral Sensors

This paper proposes an approach for the detection of changes in multitemporal Very High Resolution (VHR) optical images acquired by different multispectral sensors. The proposed approach, which is inspired by a recent framework developed to support the design of change-detection systems for single-sensor VHR remote sensing images, addresses and integrates in the general approach a strategy to effectively deal with multisensor information, i.e., to perform change detection between VHR images acquired by different multispectral sensors on two dates. This is achieved by the definition of procedures for the homogenization of radiometric, spectral and geometric image properties. These procedures map images into a common feature space where the information acquired by different multispectral sensors becomes comparable across time. Although the approach is general, here we optimize it for the detection of changes in vegetation and urban areas by employing features based on linear transformations (Tasseled Caps and Orthogonal Equations), which are shown to be effective for representing the multisensor information in a homogeneous physical way irrespectively of the considered sensor. Experiments on multitemporal images acquired by different VHR satellite systems (i.e., QuickBird, WorldView-2 and GeoEye-1) confirm the effectiveness of the proposed approach.

Earthquake damage mapping: An overall assessment of ground surveys and VHR image change detection after L'Aquila 2009 earthquake

Earth Observation (EO) data are used to map mostly affected urban areas after an earthquake generally exploiting change detection techniques applied at pixel scale. However, Civil Protection Services require damage assessment of each building according to a well-established scale to manage rescue operations and to estimate the economic losses.Considering the earthquake that hit L'Aquila city (Italy) on April 6, 2009, this work assess the feasibility of producing damage maps at the scale of single building from Very High Resolution (VHR) optical images collected before and after the seismic event. We considered the European Macroseismic Scale 1998 (EMS-98) and assessed the possibility to discriminate between collapsed or heavy damaged buildings (damage grade DG equal to 5 in the EMS-98 scale) and less damaged or undamaged buildings (DG < 5 in the EMS-98).The proposed approach relies on a pre-existing urban map to identify image objects corresponding to building footprints. The image analysis is carried out according to many different parameters with the objective of assessing their effectiveness in singling out changes associated to the building collapse. Features describing texture and colour changes, as well statistical similarity and correlation descriptors, such as the Kullbach Leibler Distance and the Mutual Information, were included in our analysis. Two supervised classification approaches, respectively, based on the use of the Bayesian Maximum A Posteriori (MAP) criterion and on Support Vector Machines (SVM), were compared. In our experiment, we considered the whole L'Aquila historical centre comparing classification results with the ground survey performed by the Istituto Nazionale di Geofisica e Vulcanologia (INGV).The work represents one of the first attempt to detect damage at the scale of single building, validated against an extensive ground survey. It addresses methodological aspects, highlighting the potential of textural features computed at object scale and SVMs, and discuss potential and limitations of EO in this field compared to ground surveys.

Fully Convolutional Networks for Semantic Segmentation of Very High Resolution Remotely Sensed Images Combined With DSM

Recently, approaches based on fully convolutional networks (FCN) have achieved state-of-the-art performance in the semantic segmentation of very high resolution (VHR) remotely sensed images. One central issue in this method is the loss of detailed information due to downsampling operations in FCN. To solve this problem, we introduce the maximum fusion strategy that effectively combines semantic information from deep layers and detailed information from shallow layers. Furthermore, this letter develops a powerful backend to enhance the result of FCN by leveraging the digital surface model, which provides height information for VHR images. The proposed semantic segmentation scheme has achieved an overall accuracy of 90.6% on the ISPRS Vaihingen benchmark.

Block adjustment with airbrone SAR very high-resolution images using trajectory constraints

ABSTRACTFor mapping a large area with airbrone synthetic aperture radar (SAR), very high-resolution (VHR) images using radargrammtry, block adjustment, where strips and blocks of images are processed together, is a preferred method for geometric processing with the advantage of reducing the needed number of ground control points (GCPs). However, the weak intersection geometry between adjacent images from the same strip conflicts with the solvability of the iterative least squares adjustment. In order to overcome this challenge, this article focuses on block adjustment with airbrone SAR VHR images using r-D model combined with trajectory constraints from azimuth tie points (ATPs). Initially, r-D model is briefly introduced. Then, the trajectory constraint equations are proposed and formulated with ATPs. Finally, an improved solution method is applied to increase the computational efficiency of the least squares process. Experiments are carried out using Chinese airbrone SAR VHR images over Shanxi Yanliang, China. The results show that the proposed method can effectively improve the accuracy of block adjustment in both range and elevation direction, with the root mean square error reduced from 0.440 to 0.253 m in range direction and from 0.501 to 0.256 m in elevation direction, respectively. The effects of the distribution and number of GCPs are also investigated, based on which some rules-of-thumb are derived.

Identifying Generalizable Image Segmentation Parameters for Urban Land Cover Mapping through Meta-Analysis and Regression Tree Modeling

The advent of very high resolution (VHR) satellite imagery and the development of Geographic Object-Based Image Analysis (GEOBIA) have led to many new opportunities for fine-scale land cover mapping, especially in urban areas. Image segmentation is an important step in the GEOBIA framework, so great time/effort is often spent to ensure that computer-generated image segments closely match real-world objects of interest. In the remote sensing community, segmentation is frequently performed using the multiresolution segmentation (MRS) algorithm, which is tuned through three user-defined parameters (the scale, shape/color, and compactness/smoothness parameters). The scale parameter (SP) is the most important parameter and governs the average size of generated image segments. Existing automatic methods to determine suitable SPs for segmentation are scene-specific and often computationally intensive, so an approach to estimating appropriate SPs that is generalizable (i.e., not scene-specific) could speed up the GEOBIA workflow considerably. In this study, we attempted to identify generalizable SPs for five common urban land cover types (buildings, vegetation, roads, bare soil, and water) through meta-analysis and nonlinear regression tree (RT) modeling. First, we performed a literature search of recent studies that employed GEOBIA for urban land cover mapping and extracted the MRS parameters used, the image properties (i.e., spatial and radiometric resolutions), and the land cover classes mapped. Using this data extracted from the literature, we constructed RT models for each land cover class to predict suitable SP values based on the: image spatial resolution, image radiometric resolution, shape/color parameter, and compactness/smoothness parameter. Based on a visual and quantitative analysis of results, we found that for all land cover classes except water, relatively accurate SPs could be identified using our RT modeling results. The main advantage of our approach over existing SP selection approaches is that our RT model results are not scene-specific, so they can be used to quickly identify suitable SPs in other VHR images.

Fine Registration for VHR Images Based on Superpixel Registration-Noise Estimation

Local nonlinear geometric distortion is problematic in the registration of very high-resolution (VHR) images. In the standard registration approach, the precision of control points generated from salient feature matching cannot be guaranteed. This letter introduces a novel superpixel registration-noise (RN) estimation method based on a two-step fine registration technique that can be estimate and mitigate the local residual misalignments in VHR images. The first step employs superpixel sparse representation and multiple displacement analysis to estimate RN information of the preregistered image. The second step optimizes the control points obtained in preregistration by combining the RN information and gross error information, and finally fine registers the input image by employing local rectification. The experiments using two data sets generated from Chinese GF2, GF1, and ZY3 satellites are discussed in this letter, and the promising results verify the effectiveness of the proposed new method.

Disparity-Based Generation of Line-of-Sight DSM for Image-Elevation Co-Registration to Support Building Detection in Off-Nadir VHR Satellite Images

The integration of optical images and elevation data is of great importance for 3D-assisted mapping applications. Very high resolution (VHR) satellite images provide ideal geo-data for mapping building information. Since buildings are inherently elevated objects, these images need to be co-registered with their elevation data for reliable building detection results. However, accurate co-registration is extremely difficult for off-nadir VHR images acquired over dense urban areas. Therefore, this research proposes a Disparity-Based Elevation Co-Registration (DECR) method for generating a Line-of-Sight Digital Surface Model (LoS-DSM) to efficiently achieve image-elevation data co-registration with pixel-level accuracy. Relative to the traditional photogrammetric approach, the RMSE value of the derived elevations is found to be less than 2 pixels. The applicability of the DECR method is demonstrated through elevation-based building detection (EBD) in a challenging dense urban area. The quality of the detection result is found to be more than 90%. Additionally, the detected objects were geo-referenced successfully to their correct ground locations to allow direct integration with other maps. In comparison to the original LoS-DSM development algorithm, the DECR algorithm is more efficient by reducing the calculation steps, preserving the co-registration accuracy, and minimizing the need for elevation normalization in dense urban areas.

A Shadow-Overlapping Algorithm for Estimating Building Heights From VHR Satellite Images

Building height is a key geometric attribute for generating 3-D building models. We propose a novel four-stage approach for automated estimation of building heights from their shadows in very high resolution (VHR) multispectral images. First, a building’s actual shadow regions are detected by applying ratio-band algorithm to the VHR image. Second, 2-D building footprint geometries are identified using graph theory and morphological fuzzy processing techniques. Third, artificial shadow regions are simulated using the identified building footprint and solar information in the image metadata at predefined height increments. Finally, the difference between the actual and simulated shadow regions at every height increment is computed using Jaccard similarity coefficient. The estimated building height corresponds to the height of the simulated shadow region that resulted in the maximum value for Jaccard index. The algorithm is tested on seven urban sites in Cardiff, U.K. with various levels of morphological complexity. Our method outperforms the past attempts, and the mean error is reduced by at least 21%.

Tile-Based Semisupervised Classification of Large-Scale VHR Remote Sensing Images

This paper deals with the problem of the classification of large-scale very high-resolution (VHR) remote sensing (RS) images in a semisupervised scenario, where we have a limited training set (less than ten training samples per class). Typical pixel-based classification methods are unfeasible for large-scale VHR images. Thus, as a practical and efficient solution, we propose to subdivide the large image into a grid of tiles and then classify the tiles instead of classifying pixels. Our proposed method uses the power of a pretrained convolutional neural network (CNN) to first extract descriptive features from each tile. Next, a neural network classifier (composed of 2 fully connected layers) is trained in a semisupervised fashion and used to classify all remaining tiles in the image. This basically presents a coarse classification of the image, which is sufficient for many RS application. The second contribution deals with the employment of the semisupervised learning to improve the classification accuracy. We present a novel semisupervised approach which exploits both the spectral and spatial relationships embedded in the remaining unlabelled tiles. In particular, we embed a spectral graph Laplacian in the hidden layer of the neural network. In addition, we apply regularization of the output labels using a spatial graph Laplacian and the random Walker algorithm. Experimental results obtained by testing the method on two large-scale images acquired by the IKONOS2 sensor reveal promising capabilities of this method in terms of classification accuracy even with less than ten training samples per class.

Deep Fully Convolutional Networks for the Detection of Informal Settlements in VHR Images

This letter investigates fully convolutional networks (FCNs) for the detection of informal settlements in very high resolution (VHR) satellite images. Informal settlements or slums are proliferating in developing countries and their detection and classification provides vital information for decision making and planning urban upgrading processes. Distinguishing different urban structures in VHR images is challenging because of the abstract semantic definition of the classes as opposed to the separation of standard land-cover classes. This task requires extraction of texture and spatial features. To this aim, we introduce deep FCNs to perform pixel-wise image labeling by automatically learning a higher level representation of the data. Deep FCNs can learn a hierarchy of features associated to increasing levels of abstraction, from raw pixel values to edges and corners up to complex spatial patterns. We present a deep FCN using dilated convolutions of increasing spatial support. It is capable of learning informative features capturing long-range pixel dependencies while keeping a limited number of network parameters. Experiments carried out on a Quickbird image acquired over the city of Dar es Salaam, Tanzania, show that the proposed FCN outperforms state-of-the-art convolutional networks. Moreover, the computational cost of the proposed technique is significantly lower than standard patch-based architectures.

Unsupervised-Restricted Deconvolutional Neural Network for Very High Resolution Remote-Sensing Image Classification

As the acquisition of very high resolution (VHR) satellite images becomes easier owing to technological advancements, ever more stringent requirements are being imposed on automatic image interpretation. Moreover, per-pixel classification has become the focus of research interests in this regard. However, the efficient and effective processing and the interpretation of VHR satellite images remain a critical task. Convolutional neural networks (CNNs) have recently been applied to VHR satellite images with considerable success. However, the prevalent CNN models accept input data of fixed sizes and train the classifier using features extracted directly from the convolutional stages or the fully connected layers, which cannot yield pixel-to-pixel classifications. Moreover, training a CNN model requires large amounts of labeled reference data. These are challenging to obtain because per-pixel labeled VHR satellite images are not open access. In this paper, we propose a framework called the unsupervised-restricted deconvolutional neural network (URDNN). It can solve these problems by learning an end-to-end and pixel-to-pixel classification and handling a VHR classification using a fully convolutional network and a small number of labeled pixels. In URDNN, supervised learning is always under the restriction of unsupervised learning, which serves to constrain and aid supervised training in learning more generalized and abstract feature. To some degree, it will try to reduce the problems of overfitting and undertraining, which arise from the scarcity of labeled training data, and to gain better classification results using fewer training samples. It improves the generality of the classification model. We tested the proposed URDNN on images from the Geoeye and Quickbird sensors and obtained satisfactory results with the highest overall accuracy (OA) achieved as 0.977 and 0.989, respectively. Experiments showed that the combined effects of additional kernels and stages may have produced better results, and two-stage URDNN consistently produced a more stable result. We compared URDNN with four methods and found that with a small ratio of selected labeled data items, it yielded the highest and most stable results, whereas the accuracy values of the other methods quickly decreased. For some categories with fewer training pixels, accuracy for categories from other methods was considerably worse than that in URDNN, with the largest difference reaching almost 10%. Hence, the proposed URDNN can successfully handle the VHR image classification using a small number of labeled pixels. Furthermore, it is more effective than state-of-the-art methods.