Search

Linear spectral unmixing is the practice of decomposing the mixed pixel into a linear combination of the constituent endmembers and the estimated abundances. This paper focuses on unsupervised spectral unmixing where the endmembers are unknown a priori. Conventional approaches use either geometrical- or statistical-based approaches. In this paper, we address the challenges of spectral unmixing with unsupervised deep learning models, in specific, the autoencoder models, where the decoder serves as the endmembers and the hidden layer output serves as the abundances. In several recent attempts, part-based autoencoders have been designed to solve the unsupervised spectral unmixing problem. However, the performance has not been satisfactory. In this paper, we first discuss some important findings we make on issues with part-based autoencoders. By proof of counterexample, we show that all existing part-based autoencoder networks with nonnegative and tied encoder and decoder are inherently defective by making these inappropriate assumptions on the network structure. As a result, they are not suitable for solving the spectral unmixing problem. We propose a so-called untied denoising autoencoder with sparsity, in which the encoder and decoder of the network are independent, and only the decoder of the network is enforced to be nonnegative. Furthermore, we make two critical additions to the network design. First, since denoising is an essential step for spectral unmixing, we propose to incorporate the denoising capacity into the network optimization in the format of a denoising constraint rather than cascading another denoising preprocessor in order to avoid the introduction of additional reconstruction error. Second, to be more robust to the inaccurate estimation of a number of endmembers, we adopt an $l_{21}$ -norm on the encoder of the network to reduce the redundant endmembers while decreasing the reconstruction error simultaneously. The experimental results demonstrate that the proposed approach outperforms several state-of-the-art methods, especially for highly noisy data. [ABSTRACT FROM AUTHOR]

Deep models with multiple layers have demonstrated their potential in learning abstract and invariant features for better representation and classification of remote sensing images. Moreover, metric learning (ML) is usually introduced into the deep models to further increase the discrimination of deep representations. However, the usual deep ML methods treat the training samples in each training batch in the stochastic gradient descent-based learning procedure independently, and thus, they neglect the important contextual (structural) information in the training samples. In this paper, we first introduce deep structural ML (DSML) into the literature of remote sensing scene classification and specifically capture and use the structural information during the training on the remote sensing images. Further analysis demonstrates that DSML usually makes many learned metric parameters similar. This similarity leads to obvious model redundancy and thus decreases the representational ability of the model. To address this problem, this paper proposes a new diversity-promoting DSML (D-DSML) method by regularizing the learning procedure by a diversity-promoting prior over the parameter factors. The proposed D-DSML encourages the parameter factors to be uncorrelated, such that each factor can model unique information, and thus, the model's description ability and classification performance would be significantly improved. Experiments over six real-world remote sensing scene data sets demonstrate that the proposed method obtains much better results than those obtained by the original deep models and has comparable or even better performances when compared with state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Change detection has been a hotspot in the remote sensing technology for a long time. With the increasing availability of multi-temporal remote sensing images, numerous change detection algorithms have been proposed. Among these methods, image transformation methods with feature extraction and mapping could effectively highlight the changed information and thus has a better change detection performance. However, the changes of multi-temporal images are usually complex, and the existing methods are not effective enough. In recent years, the deep network has shown its brilliant performance in many fields, including feature extraction and projection. Therefore, in this paper, based on the deep network and slow feature analysis (SFA) theory, we proposed a new change detection algorithm for multi-temporal remotes sensing images called deep SFA (DSFA). In the DSFA model, two symmetric deep networks are utilized for projecting the input data of bi-temporal imagery. Then, the SFA module is deployed to suppress the unchanged components and highlight the changed components of the transformed features. The change vector analysis pre-detection is employed to find unchanged pixels with high confidence as training samples. Finally, the change intensity is calculated with chi-square distance and the changes are determined by threshold algorithms. The experiments are performed on two real-world data sets and a public hyperspectral data set. The visual comparison and the quantitative evaluation have shown that DSFA could outperform the other state-of-the-art algorithms, including other SFA-based and deep learning methods. [ABSTRACT FROM AUTHOR]

Accurate and robust detection of multi-class objects in optical remote sensing images is essential to many real-world applications, such as urban planning, traffic control, searching, and rescuing. However, the state-of-the-art object detection techniques designed for images captured using ground-level sensors usually experience a sharp performance drop when directly applied to remote sensing images, largely due to the object appearance differences in remote sensing images in terms of sparse texture, low contrast, arbitrary orientations, and large-scale variations. This paper presents a novel object detection network [(context-aware detection network (CAD-Net)] that exploits attention-modulated features as well as global and local contexts to address the new challenges in detecting objects from remote sensing images. The proposed CAD-Net learns global and local contexts of objects by capturing their correlations with the global scene (at scene level) and the local neighboring objects or features (at object level), respectively. In addition, it designs a spatial-and-scale-aware attention module that guides the network to focus on more informative regions and features as well as more appropriate feature scales. Experiments over two publicly available object detection data sets for remote sensing images demonstrate that the proposed CAD-Net achieves superior detection performance. The implementation codes will be made publicly available for facilitating future works. [ABSTRACT FROM AUTHOR]

While deep learning technologies for computer vision have developed rapidly since 2012, modeling of remote sensing systems has remained focused around human vision. In particular, remote sensing systems are usually constructed to optimize sensing cost-quality tradeoffs with respect to human image interpretability. While some recent studies have explored remote sensing system design as a function of simple computer vision algorithm performance, there has been little work relating this design to the state of the art in computer vision: deep learning with convolutional neural networks. We develop experimental systems to conduct this analysis, showing results with modern deep learning algorithms and recent overhead image data. Our results are compared to standard image quality measurements based on human visual perception, and we conclude not only that machine and human interpretability differ significantly but also that computer vision performance is largely self-consistent across a range of disparate conditions. This paper is presented as a cornerstone for a new generation of sensor design systems that focus on computer algorithm performance instead of human visual perception. [ABSTRACT FROM AUTHOR]

Frequency filtering is widely used in routine processing of seismic data to improve the signal-to-noise ratio (SNR) of recorded signals and by doing so to improve subsequent analyses. In this paper, we develop a new denoising/decomposition method, DeepDenoiser, based on a deep neural network. This network is able to simultaneously learn a sparse representation of data in the time–frequency domain and a non-linear function that maps this representation into masks that decompose input data into a signal of interest and noise (defined as any non-seismic signal). We show that DeepDenoiser achieves impressive denoising of seismic signals even when the signal and noise share a common frequency band. Because the noise statistics are automatically learned from data and require no assumptions, our method properly handles white noise, a variety of colored noise, and non-earthquake signals. DeepDenoiser can significantly improve the SNR with minimal changes in the waveform shape of interest, even in the presence of high noise levels. We demonstrate the effect of our method on improving earthquake detection. There are clear applications of DeepDenoiser to seismic imaging, micro-seismic monitoring, and preprocessing of ambient noise data. We also note that the potential applications of our approach are not limited to these applications or even to earthquake data and that our approach can be adapted to diverse signals and applications in other settings. [ABSTRACT FROM AUTHOR]

Recently, deep neural networks have received intense interests in polarimetric synthetic aperture radar (PolSAR) image classification. However, its success is subject to the availability of large amounts of annotated data which require great efforts of experienced human annotators. Aiming at improving the classification performance with greatly reduced annotation cost, this paper presents an active deep learning approach for minimally supervised PolSAR image classification, which integrates active learning and fine-tuned convolutional neural network (CNN) into a principled framework. Starting from a CNN trained using a very limited number of labeled pixels, we iteratively and actively select the most informative candidates for annotation, and incrementally fine-tune the CNN by incorporating the newly annotated pixels. Moreover, to boost the performance and robustness of the proposed method, we employ Markov random field (MRF) to enforce class label smoothness, and data augmentation technique to enlarge the training set. We conducted extensive experiments on four real benchmark PolSAR images, and experiments demonstrated that our approach achieved state-of-the-art classification results with significantly reduced annotation cost. [ABSTRACT FROM AUTHOR]

For hyperspectral remote sensing image (HSI) classification, the learning process of deep neural networks has been progressively advanced in depth, but the fine features are often largely lost or even disappear in the process of depth transfer. With the increase in feature aggregation and connectivity, the complexity of the network and the training parameters increases greatly, requiring more training time. This paper proposed a multi-scale dense network (MSDN) for HSI classification that made full use of different scale information in the network structure and combined scale information throughout the network. It implemented feature extraction of HSIs in two dimensions, including the features at fine and coarse levels. In the horizontal direction, it considered the deep extraction of HSI features, and the 3-D dense connection structure was used for aggregating features at different levels. In the vertical direction, scale information was considered, and three-scale feature maps at low, middle, and high levels were generated based on the first layer of the network. The MSDN used stride convolution for downsampling and combined feature information at different scale levels. The MSDN extracted features along the diagonal line. The network implemented the reconstruction of deep feature extraction and multi-scale fusion for HSI classification. The MSDN model performed well on representative HSI datasets, namely, the Indian Pines, Pavia University, Salinas, Botswana, and Kennedy Space Center datasets. It improved the training speed and accuracy for HSI classification and especially improved the convergence speed, which effectively saved computing resources and had high stability. [ABSTRACT FROM AUTHOR]

In order to better interpret polarimetric synthetic aperture radar (PolSAR) images, many scholars tend to do target decomposition for PolSAR images and utilize the obtained features to perform subsequent classification. These target decomposition features play an important role in terrain classification but completely utilizing them produces a high computational complexity. Furthermore, some features have a negative impact on the classification task. Therefore, selecting the appropriate amount of high-quality features is of great significance to the classification task. In this paper, we propose a convolutional neural network (CNN)-based feature selection algorithm for PolSAR image classification. First, we design a 1-D CNN for feature selection, then train the designed network with all the decomposition features to obtain a trained model. Second, the Kullback–Leibler distance (KLD) between different features is utilized as a standard to select feature subsets. Third, feature subsets with excellent performance form the final results. Due to the special structure of the 1-D CNN, repetitively training model is avoided when the input changes. Different from traditional feature selection methods, our method considers the performance of features combination rather than single feature contribution. To this end, the feature subsets selected by the proposed method are more useful to the classification task. Innovatively introducing KLD in the selection stage avoids random selection and improves the selection efficiency. Finally, we validate the performance of selected feature subsets in traditional and deep learning classification frameworks. Experiments demonstrate that features selected by the proposed method have a good performance comparing with others on three real PolSAR data sets. [ABSTRACT FROM AUTHOR]

Point cloud classification is one of the most important and time-consuming stages of airborne LiDAR (Light Detection and Ranging) data processing, playing a key role in the generation of cartographic products. This paper describes an innovative algorithm to perform LiDAR point-cloud classification, which relies on Convolutional Neural Networks (CNNs) and takes advantage of full-waveform data registered by modern laser scanners. The proposed method consists of two steps. First, a simple CNN is used to preprocess each waveform, providing a compact representation of the data. By exploiting the coordinates of the points associated with the waveforms, output vectors generated by the first CNN are then mapped into an image that is subsequently segmented by a Fully Convolutional Network (FCN): a label is assigned to each pixel and, consequently, to the point falling in the pixel. In this way, spatial positions and geometrical relationships between neighboring data are taken into account. These particular architectures allow to accurately identify even challenging classes such as power line and transmission tower. [ABSTRACT FROM AUTHOR]

The automatic classification of 3-D point clouds is publicly known as a challenging task in a complex road environment. Specifically, each point is automatically classified into a unique category label, and then, the labels are used as clues for semantic analysis and scene recognition. Instead of heuristically extracting handcrafted features in traditional methods to classify all points, we put forward an end-to-end octree-based fully convolutional network (FCN) to classify 3-D point clouds in an urban road environment. There are four contributions in this paper. The first is that the integration and comprehensive uses of OctNet and FCN greatly decrease the computing time and memory demands compared with a dense 3-D convolutional neural network (CNN). The second is that the octree-based network is strengthened by means of modifying the cross-entropy loss function to solve the problems of an unbalanced category distribution. The third is that an Inception-ResNet block is united with our network, which enables our 3-D CNN to effectively learn how to classify scenes containing objects at multiple scales and improve classification accuracy. The last is that an open source data set (HuangshiRoad data set) with ten different classes is introduced for 3-D point cloud classification. Three representative data sets [Semantic3D, WHU_MLS (blocks I and II), and HuangshiRoad] with different covered areas and numbers of points and classes are selected to evaluate our proposed method. The experimental results show that the overall classification accuracy is appreciable, with 89.4% for Semantic3D, 82.9% for WHU_MLS block I, 91.4% for WHU_MLS block II, and 94% for HuangshiRoad. Our deep learning approach can efficiently classify 3-D dense point clouds in an urban road environment measured by a mobile laser scanning (MLS) system or static LiDAR. [ABSTRACT FROM AUTHOR]

Road information extraction based on aerial images is a critical task for many applications, and it has attracted considerable attention from researchers in the field of remote sensing. The problem is mainly composed of two subtasks, namely, road detection and centerline extraction. Most of the previous studies rely on multistage-based learning methods to solve the problem. However, these approaches may suffer from the well-known problem of propagation errors. In this paper, we propose a novel deep learning model, recurrent convolution neural network U-Net (RCNN-UNet), to tackle the aforementioned problem. Our proposed RCNN-UNet has three distinct advantages. First, the end-to-end deep learning scheme eliminates the propagation errors. Second, a carefully designed RCNN unit is leveraged to build our deep learning architecture, which can better exploit the spatial context and the rich low-level visual features. Thereby, it alleviates the detection problems caused by noises, occlusions, and complex backgrounds of roads. Third, as the tasks of road detection and centerline extraction are strongly correlated, a multitask learning scheme is designed so that two predictors can be simultaneously trained to improve both effectiveness and efficiency. Extensive experiments were carried out based on two publicly available benchmark data sets, and nine state-of-the-art baselines were used in a comparative evaluation. Our experimental results demonstrate the superiority of the proposed RCNN-UNet model for both the road detection and the centerline extraction tasks. [ABSTRACT FROM AUTHOR]

The shortage of training samples remains one of the main obstacles in applying the neural networks to the hyperspectral images classification. To fuse the spatial and spectral information, pixel patches are often utilized to train a model, which may further aggregate this problem. In the existing works, an artificial neural network (ANN) model supervised by centerloss (ANNC) was introduced. Training merely with spectral information, the ANNC yields discriminative spectral features suitable for the subsequent classification tasks. In this paper, we propose a novel convolutional neural network (CNN)-based spatial feature fusion (CSFF) algorithm, which allows a smart integration of spatial information to the spectral features extracted by ANNC. As a critical part of CSFF, a CNN-based discriminant model is introduced to estimate whether two pixels belong to the same class. At the testing stage, by applying the discriminant model to the pixel pairs generated by a test pixel and each of its neighbors, the local structure is estimated and represented as a customized convolutional kernel. The spectral–spatial feature is generated by a convolutional operation between the estimated kernel and the corresponding spectral features within a local region. The final label is determined by classifying the resulting spectral–spatial feature. Without increasing the number of training samples or involving pixel patches at the training stage, the CSFF framework achieves the state of the art by declining 20%–50% classification failures in experiments on three well-known hyperspectral images. [ABSTRACT FROM AUTHOR]

Hyperspectral image (HSI) classification is a core task in the remote sensing community, and recently, deep learning-based methods have shown their capability of accurate classification of HSIs. Among the deep learning-based methods, deep convolutional neural networks (CNNs) have been widely used for the HSI classification. In order to obtain a good classification performance, substantial efforts are required to design a proper deep learning architecture. Furthermore, the manually designed architecture may not fit a specific data set very well. In this paper, the idea of automatic CNN for the HSI classification is proposed for the first time. First, a number of operations, including convolution, pooling, identity, and batch normalization, are selected. Then, a gradient descent-based search algorithm is used to effectively find the optimal deep architecture that is evaluated on the validation data set. After that, the best CNN architecture is selected as the model for the HSI classification. Specifically, the automatic 1-D Auto-CNN and 3-D Auto-CNN are used as spectral and spectral–spatial HSI classifiers, respectively. Furthermore, the cutout is introduced as a regularization technique for the HSI spectral–spatial classification to further improve the classification accuracy. The experiments on four widely used hyperspectral data sets (i.e., Salinas, Pavia University, Kennedy Space Center, and Indiana Pines) show that the automatically designed data-dependent CNNs obtain competitive classification accuracy compared with the state-of-the-art methods. In addition, the automatic design of the deep learning architecture opens a new window for future research, showing the huge potential of using neural architectures’ optimization capabilities for the accurate HSI classification. [ABSTRACT FROM AUTHOR]

This paper addresses the problem of training a deep neural network for satellite image segmentation so that it can be deployed over images whose statistics differ from those used for training. For example, in postdisaster damage assessment, the tight time constraints make it impractical to train a network from scratch for each image to be segmented. We propose a convolutional encoder–decoder network able to learn visual representations of increasing semantic level as its depth increases, allowing it to generalize over a wider range of satellite images. Then, we propose two additional methods to improve the network performance over each specific image to be segmented. First, we observe that updating the batch normalization layers’ statistics over the target image improves the network performance without human intervention. Second, we show that refining a trained network over a few samples of the image boosts the network performance with minimal human intervention. We evaluate our architecture over three data sets of satellite images, showing the state-of-the-art performance in binary segmentation of previously unseen images and competitive performance with respect to more complex techniques in a multiclass segmentation task. [ABSTRACT FROM AUTHOR]

In the field of spatial–spectral fusion, the variational model-based methods and the deep learning (DL)-based methods are state-of-the-art approaches. This paper presents a fusion method that combines the deep neural network with a variational model for the most common case of spatial–spectral fusion: panchromatic (PAN)/multispectral (MS) fusion. Specifically, a deep residual convolutional neural network (CNN) is first trained to learn the gradient features of the high spatial resolution multispectral image (HR-MS). The image observation variational models are then formulated to describe the relationships of the ideal fused image, the observed low spatial resolution multispectral image (LR-MS) image, and the gradient priors learned before. Then, fusion result can then be obtained by solving the fusion variational model. Both quantitative and visual assessments on high-quality images from various sources demonstrate that the proposed fusion method is superior to all the mainstream algorithms included in the comparison, in terms of overall fusion accuracy. [ABSTRACT FROM AUTHOR]

Deep learning artificial neural network techniques can be applied for on-orbit calibration of microwave and millimeter-wave radiometer spaceborne instruments, including those for small satellites. The noise-wave model has been employed for noise characterization and validation of the proposed deep learning calibration technique for a synthetically generated Dicke-switching radiometer. The developed deep learning neural network radiometer calibrator produces high accuracy estimates of antenna temperatures from the measurements of radiometer output voltage and thermistor readings. Tests with noise-free and noisy samples of the developed model have shown that the proposed calibration method does not add any significant noise to the radiometer calibration. The performance of the proposed method does not degrade with increased nonlinearity for a radiometer, while nonlinearity is a challenging issue for conventional calibration techniques. The deep learning calibration model learns the radiometer noise characteristics from radiometer prelaunch measurements during thermal vacuum chamber testing. The neural network calibrator proposed in this paper has self-learning capability during the on-orbit operation of a radiometer that can be used to improve the performance of on-orbit calibration. The proposed technique is demonstrated by comparing the residual uncertainty of the deep learning calibration with the theoretical value. No numerical study is presented to compare the performance with conventional calibration techniques. The new method may be solely applied to calibrate the radiometer or applied along with conventional calibration techniques. [ABSTRACT FROM AUTHOR]

Aiming at improving the classification accuracy with limited numbers of labeled pixels in polarimetric synthetic aperture radar (PolSAR) image classification task, this paper presents a graph-based semisupervised deep learning model for PolSAR image classification. It models the PolSAR image as an undirected graph, where the nodes correspond to the labeled and unlabeled pixels, and the weighted edges represent similarities between the pixels. Upon the graph, we design an energy function incorporating a semisupervision term, a convolutional neural network (CNN) term, and a pairwise smoothness term. The employed CNN extracts abstract and data-driven polarimetric features and outputs class label predictions to the graph model. The semisupervision term enforces the category label constraints on the human-labeled pixels. The pairwise smoothness term encourages class label smoothness and the alignment of class label boundaries with the image edges. Starting from an initialized class label map generated based on $K$ -Wishart distribution hypothesis or superpixel segmentation of PauliRGB images, we iteratively and alternately optimize the defined energy function until it converges. We conducted experiments on two real benchmark PolSAR images, and extensive experiments demonstrated that our approach achieved the state-of-the-art results for PolSAR image classification. [ABSTRACT FROM AUTHOR]

In this paper, we propose a spectral-spatial feature extraction and classification framework based on an artificial neuron network in the context of hyperspectral imagery. With limited labeled samples, only spectral information is exploited for training and spatial context is integrated posteriorly at the testing stage. Taking advantage of recent advances in face recognition, a joint supervision symbol that combines softmax loss and center loss is adopted to train the proposed network, by which intraclass features are gathered while interclass variations are enlarged. Based on the learned architecture, the extracted spectrum-based features are classified by a center classifier. Moreover, to fuse the spectral and spatial information, an adaptive spectral-spatial center classifier is developed, where multiscale neighborhoods are considered simultaneously, and the final label is determined using an adaptive voting strategy. Finally, experimental results on three well-known data sets validate the effectiveness of the proposed methods compared with the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Deep learning has recently attracted significant attention in the field of hyperspectral images (HSIs) classification. However, the construction of an efficient deep neural network mostly relies on a large number of labeled samples being available. To address this problem, this paper proposes a unified deep network, combined with active transfer learning (TL) that can be well-trained for HSIs classification using only minimally labeled training data. More specifically, deep joint spectral–spatial feature is first extracted through hierarchical stacked sparse autoencoder (SSAE) networks. Active TL is then exploited to transfer the pretrained SSAE network and the limited training samples from the source domain to the target domain, where the SSAE network is subsequently fine-tuned using the limited labeled samples selected from both source and target domains by the corresponding active learning (AL) strategies. The advantages of our proposed method are threefold: 1) the network can be effectively trained using only limited labeled samples with the help of novel AL strategies; 2) the network is flexible and scalable enough to function across various transfer situations, including cross data set and intraimage; and 3) the learned deep joint spectral–spatial feature representation is more generic and robust than many joint spectral–spatial feature representations. Extensive comparative evaluations demonstrate that our proposed method significantly outperforms many state-of-the-art approaches, including both traditional and deep network-based methods, on three popular data sets. [ABSTRACT FROM AUTHOR]

Satellite images are widely applied in humanitarian mapping that labels buildings, roads, and so on for humanitarian aid and economic development. However, the labeling now is mostly done by volunteers. In this paper, we utilize deep learning to solve humanitarian mapping tasks of a mobile software named MapSwipe. The current deep learning techniques, e.g., convolutional neural network (CNN), can recognize ground objects from satellite images but rely on numerous labels for training for each specific task. We solve this problem by fusing multiple freely accessible crowdsourced geographic data and propose an active learning-based CNN training framework named MC-CNN to deal with the quality issues of the labels extracted from these data, including incompleteness (e.g., some kinds of object are not labeled) and heterogeneity (e.g., different spatial granularities). The method is evaluated with building mapping in South Malawi and road mapping in Guinea with level-18 satellite images provided by Bing Map and volunteered geographic information from OpenStreetMap, MapSwipe, and OsmAnd. The results based on multiple metrics, including Precision, Recall, F1 Score, and area under the receiver operating characteristic curve, show that MC-CNN can fuse the crowdsourced labels for higher prediction performance and be successfully applied in MapSwipe for humanitarian mapping with 85% labor saved and an overall accuracy of 0.86 achieved. [ABSTRACT FROM AUTHOR]

Hyperspectral image (HSI) denoising is a crucial preprocessing procedure to improve the performance of the subsequent HSI interpretation and applications. In this paper, a novel deep learning-based method for this task is proposed, by learning a nonlinear end-to-end mapping between the noisy and clean HSIs with a combined spatial–spectral deep convolutional neural network (HSID-CNN). Both the spatial and spectral information are simultaneously assigned to the proposed network. In addition, multiscale feature extraction and multilevel feature representation are, respectively, employed to capture both the multiscale spatial–spectral feature and fuse different feature representations for the final restoration. The simulated and real-data experiments demonstrate that the proposed HSID-CNN outperforms many of the mainstream methods in both the quantitative evaluation indexes, visual effects, and HSI classification accuracy. [ABSTRACT FROM AUTHOR]

Object detection and semantic segmentation are two main themes in object retrieval from high-resolution remote sensing images, which have recently achieved remarkable performance by surfing the wave of deep learning and, more notably, convolutional neural networks. In this paper, we are interested in a novel, more challenging problem of vehicle instance segmentation, which entails identifying, at a pixel level, where the vehicles appear as well as associating each pixel with a physical instance of a vehicle. In contrast, vehicle detection and semantic segmentation each only concern one of the two. We propose to tackle this problem with a semantic boundary-aware multitask learning network. More specifically, we utilize the philosophy of residual learning to construct a fully convolutional network that is capable of harnessing multilevel contextual feature representations learned from different residual blocks. We theoretically analyze and discuss why residual networks can produce better probability maps for pixelwise segmentation tasks. Then, based on this network architecture, we propose a unified multitask learning network that can simultaneously learn two complementary tasks, namely, segmenting vehicle regions and detecting semantic boundaries. The latter subproblem is helpful for differentiating “touching” vehicles that are usually not correctly separated into instances. Currently, data sets with a pixelwise annotation for vehicle extraction are the ISPRS data set and the IEEE GRSS DFC2015 data set over Zeebrugge, which specializes in a semantic segmentation. Therefore, we built a new, more challenging data set for vehicle instance segmentation, called the Busy Parking Lot Unmanned Aerial Vehicle Video data set, and we make our data set available at http://www.sipeo.bgu.tum.de/downloads so that it can be used to benchmark future vehicle instance segmentation algorithms. [ABSTRACT FROM AUTHOR]

Hyperspectral image (HSI) classification is an active and important research task driven by many practical applications. To leverage deep learning models especially convolutional neural networks (CNNs) for HSI classification, this paper proposes a simple yet effective method to extract hierarchical deep spatial feature for HSI classification by exploring the power of off-the-shelf CNN models, without any additional retraining or fine-tuning on the target data set. To obtain better classification accuracy, we further propose a unified metric learning-based framework to alternately learn discriminative spectral–spatial features, which have better representation capability and train support vector machine (SVM) classifiers. To this end, we design a new objective function that explicitly embeds a metric learning regularization term into SVM training. The metric learning regularization term is used to learn a powerful spectral–spatial feature representation by fusing spectral feature and deep spatial feature, which has small intraclass scatter but big between class separation. By transforming HSI data into new spectral–spatial feature space through CNN and metric learning, we can pull the pixels from the same class closer, while pushing the different class pixels farther away. In the experiments, we comprehensively evaluate the proposed method on three commonly used HSI benchmark data sets. State-of-the-art results are achieved when compared with the existing HSI classification methods. [ABSTRACT FROM AUTHOR]

Hyperspectral unmixing is an important processing step for many hyperspectral applications, mainly including: 1) estimation of pure spectral signatures (endmembers) and 2) estimation of the abundance of each endmember in each pixel of the image. In recent years, nonnegative matrix factorization (NMF) has been highly attractive for this purpose due to the nonnegativity constraint that is often imposed in the abundance estimation step. However, most of the existing NMF-based methods only consider the information in a single layer while neglecting the hierarchical features with hidden information. To alleviate such limitation, in this paper, we propose a new sparsity-constrained deep NMF with total variation (SDNMF-TV) technique for hyperspectral unmixing. First, by adopting the concept of deep learning, the NMF algorithm is extended to deep NMF model. The proposed model consists of pretraining stage and fine-tuning stage, where the former pretrains all factors layer by layer and the latter is used to reduce the total reconstruction error. Second, in order to exploit adequately the spectral and spatial information included in the original hyperspectral image, we enforce two constraints on the abundance matrix. Specifically, the $L_{1/2}$ constraint is adopted, since the distribution of each endmember is sparse in the 2-D space. The TV regularizer is further introduced to promote piecewise smoothness in abundance maps. For the optimization of the proposed model, multiplicative update rules are derived using the gradient descent method. The effectiveness and superiority of the SDNMF-TV algorithm are demonstrated by comparing with other unmixing methods on both synthetic and real data sets. [ABSTRACT FROM AUTHOR]

Deep learning has achieved great successes in conventional computer vision tasks. In this paper, we exploit deep learning techniques to address the hyperspectral image classification problem. In contrast to conventional computer vision tasks that only examine the spatial context, our proposed method can exploit both spatial context and spectral correlation to enhance hyperspectral image classification. In particular, we advocate four new deep learning models, namely, 2-D convolutional neural network (2-D-CNN), 3-D-CNN, recurrent 2-D CNN (R-2-D-CNN), and recurrent 3-D-CNN (R-3-D-CNN) for hyperspectral image classification. We conducted rigorous experiments based on six publicly available data sets. Through a comparative evaluation with other state-of-the-art methods, our experimental results confirm the superiority of the proposed deep learning models, especially the R-3-D-CNN and the R-2-D-CNN deep learning models. [ABSTRACT FROM AUTHOR]

A generative adversarial network (GAN) usually contains a generative network and a discriminative network in competition with each other. The GAN has shown its capability in a variety of applications. In this paper, the usefulness and effectiveness of GAN for classification of hyperspectral images (HSIs) are explored for the first time. In the proposed GAN, a convolutional neural network (CNN) is designed to discriminate the inputs and another CNN is used to generate so-called fake inputs. The aforementioned CNNs are trained together: the generative CNN tries to generate fake inputs that are as real as possible, and the discriminative CNN tries to classify the real and fake inputs. This kind of adversarial training improves the generalization capability of the discriminative CNN, which is really important when the training samples are limited. Specifically, we propose two schemes: 1) a well-designed 1D-GAN as a spectral classifier and 2) a robust 3D-GAN as a spectral–spatial classifier. Furthermore, the generated adversarial samples are used with real training samples to fine-tune the discriminative CNN, which improves the final classification performance. The proposed classifiers are carried out on three widely used hyperspectral data sets: Salinas, Indiana Pines, and Kennedy Space Center. The obtained results reveal that the proposed models provide competitive results compared to the state-of-the-art methods. In addition, the proposed GANs open new opportunities in the remote sensing community for the challenging task of HSI classification and also reveal the huge potential of GAN-based methods for the analysis of such complex and inherently nonlinear data. [ABSTRACT FROM AUTHOR]

Remote sensing image scene classification is an active and challenging task driven by many applications. More recently, with the advances of deep learning models especially convolutional neural networks (CNNs), the performance of remote sensing image scene classification has been significantly improved due to the powerful feature representations learnt through CNNs. Although great success has been obtained so far, the problems of within-class diversity and between-class similarity are still two big challenges. To address these problems, in this paper, we propose a simple but effective method to learn discriminative CNNs (D-CNNs) to boost the performance of remote sensing image scene classification. Different from the traditional CNN models that minimize only the cross entropy loss, our proposed D-CNN models are trained by optimizing a new discriminative objective function. To this end, apart from minimizing the classification error, we also explicitly impose a metric learning regularization term on the CNN features. The metric learning regularization enforces the D-CNN models to be more discriminative so that, in the new D-CNN feature spaces, the images from the same scene class are mapped closely to each other and the images of different classes are mapped as farther apart as possible. In the experiments, we comprehensively evaluate the proposed method on three publicly available benchmark data sets using three off-the-shelf CNN models. Experimental results demonstrate that our proposed D-CNN methods outperform the existing baseline methods and achieve state-of-the-art results on all three data sets. [ABSTRACT FROM AUTHOR]

Most of the existing deep-learning-based methods are difficult to effectively deal with the challenges faced for geospatial object detection such as rotation variations and appearance ambiguity. To address these problems, this paper proposes a novel deep-learning-based object detection framework including region proposal network (RPN) and local-contextual feature fusion network designed for remote sensing images. Specifically, the RPN includes additional multiangle anchors besides the conventional multiscale and multiaspect-ratio ones, and thus can deal with the multiangle and multiscale characteristics of geospatial objects. To address the appearance ambiguity problem, we propose a double-channel feature fusion network that can learn local and contextual properties along two independent pathways. The two kinds of features are later combined in the final layers of processing in order to form a powerful joint representation. Comprehensive evaluations on a publicly available ten-class object detection data set demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

The reconstruction of urban buildings from large-scale airborne laser scanning point clouds is an important research topic in the geoscience field. Large-scale urban scenes usually contain a large number of object categories and many overlapped or closely neighboring objects, which poses great challenges for classifying and modeling buildings from these data sets. In this paper, we propose a deep reinforcement learning framework that integrates a 3-D convolutional neural network, a deep Q-network, and a residual recurrent neural network for the efficient semantic parsing of large-scale 3-D point clouds. The proposed framework provides an end-to-end automatic processing method that maps the raw point cloud to the classification results of the given categories. After obtaining the building classes, we utilize an edge-aware resampling algorithm to consolidate the point set with noise-free normals and clean preservation of sharp features. Finally, 2.5-D dual contouring, which is a data-driven approach, is introduced to generate urban building models from the consolidated point clouds. Our method can generate lightweight building models with arbitrarily shaped roofs while preserving the verticality of connecting walls. [ABSTRACT FROM AUTHOR]

It is a feasible and promising way to utilize deep neural networks to learn and extract valuable features from synthetic aperture radar (SAR) images for SAR automatic target recognition (ATR). However, it is too difficult to effectively train the deep neural networks with limited raw SAR images. In this paper, we propose a new approach to do SAR ATR, in which a multiview deep learning framework was employed. Based on the multiview SAR ATR pattern, we first present a flexible mean to generate adequate multiview SAR data, which can guarantee a large amount of inputs for network training without needing many raw SAR images. Then, a unique deep convolutional neural network containing a parallel network topology with multiple inputs is adopted. The features of input SAR images from different views will be learned by the proposed network layer by layer; meanwhile, the learned features from the distinct views are fused in different layers progressively. Therefore, the proposed framework is able to achieve a superior recognition performance, and requires only a small number of raw SAR images for network training samples generation. Experimental results have shown the superiority of the proposed framework based on the Moving and Stationary Target Acquisition and Recognition data set. [ABSTRACT FROM AUTHOR]

As a list of remotely sensed data sources is available, how to efficiently exploit useful information from multisource data for better Earth observation becomes an interesting but challenging problem. In this paper, the classification fusion of hyperspectral imagery (HSI) and data from other multiple sensors, such as light detection and ranging (LiDAR) data, is investigated with the state-of-the-art deep learning, named the two-branch convolution neural network (CNN). More specific, a two-tunnel CNN framework is first developed to extract spectral-spatial features from HSI; besides, the CNN with cascade block is designed for feature extraction from LiDAR or highresolution visual image. In the feature fusion stage, the spatial and spectral features of HSI are first integrated in a dual-tunnel branch, and then combined with other data features extracted from a cascade network. Experimental results based on several multisource data demonstrate the proposed two-branch CNN that can achieve more excellent classification performance than some existing methods. [ABSTRACT FROM AUTHOR]

Following the great success of deep convolutional neural networks (CNNs) in computer vision, this paper proposes a complex-valued CNN (CV-CNN) specifically for synthetic aperture radar (SAR) image interpretation. It utilizes both amplitude and phase information of complex SAR imagery. All elements of CNN including input-output layer, convolution layer, activation function, and pooling layer are extended to the complex domain. Moreover, a complex backpropagation algorithm based on stochastic gradient descent is derived for CV-CNN training. The proposed CV-CNN is then tested on the typical polarimetric SAR image classification task which classifies each pixel into known terrain types via supervised training. Experiments with the benchmark data sets of Flevoland and Oberpfaffenhofen show that the classification error can be further reduced if employing CV-CNN instead of conventional real-valued CNN with the same degrees of freedom. The performance of CV-CNN is comparable to that of existing state-of-the-art methods in terms of overall classification accuracy. [ABSTRACT FROM PUBLISHER]

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. [ABSTRACT FROM PUBLISHER]

Land cover change monitoring is an important task from the perspective of regional resource monitoring, disaster management, land development, and environmental planning. In this paper, we analyze imagery data from remote sensing satellites to detect forest cover changes over a period of 29 years (1987–2015). Since the original data are severely incomplete and contaminated with artifacts, we first devise a spatiotemporal inpainting mechanism to recover the missing surface reflectance information. The spatial filling process makes use of the available data of the nearby temporal instances followed by a sparse encoding-based reconstruction. We formulate the change detection task as a region classification problem. We build a multiresolution profile (MRP) of the target area and generate a candidate set of bounding-box proposals that enclose potential change regions. In contrast to existing methods that use handcrafted features, we automatically learn region representations using a deep neural network in a data-driven fashion. Based on these highly discriminative representations, we determine forest changes and predict their onset and offset timings by labeling the candidate set of proposals. Our approach achieves the state-of-the-art average patch classification rate of 91.6% (an improvement of ~16%) and the mean onset/offset prediction error of 4.9 months (an error reduction of five months) compared with a strong baseline. We also qualitatively analyze the detected changes in the unlabeled image regions, which demonstrate that the proposed forest change detection approach is scalable to new regions. [ABSTRACT FROM PUBLISHER]

This paper proposes a new framework for deep learning that has been particularly tailored for hyperspectral image classification. We learn multiple levels of dictionaries in a robust fashion. The last layer is discriminative that learns a linear classifier. The training proceeds greedily; at a time, a single level of dictionary is learned and the coefficients used to train the next level. The coefficients from the final level are used for classification. Robustness is incorporated by minimizing the absolute deviations instead of the more popular Euclidean norm. The inbuilt robustness helps combat mixed noise (Gaussian and sparse) present in hyperspectral images. Results show that our proposed techniques outperform all other deep learning methods—deep belief network, stacked autoencoder, and convolutional neural network. The experiments have been carried out on both benchmark deep learning data sets (MNIST, CIFAR-10, and Street View House Numbers) as well as on real hyperspectral imaging data sets. [ABSTRACT FROM PUBLISHER]

Deep learning based land cover classification algorithms have recently been proposed in the literature. In hyperspectral images (HSIs), they face the challenges of large dimensionality, spatial variability of spectral signatures, and scarcity of labeled data. In this paper, we propose an end-to-end deep learning architecture that extracts band specific spectral-spatial features and performs land cover classification. The architecture has fewer independent connection weights and thus requires fewer training samples. The method is found to outperform the highest reported accuracies on popular HSI data sets. [ABSTRACT FROM PUBLISHER]

In recent years, vector-based machine learning algorithms, such as random forests, support vector machines, and 1-D convolutional neural networks, have shown promising results in hyperspectral image classification. Such methodologies, nevertheless, can lead to information loss in representing hyperspectral pixels, which intrinsically have a sequence-based data structure. A recurrent neural network (RNN), an important branch of the deep learning family, is mainly designed to handle sequential data. Can sequence-based RNN be an effective method of hyperspectral image classification? In this paper, we propose a novel RNN model that can effectively analyze hyperspectral pixels as sequential data and then determine information categories via network reasoning. As far as we know, this is the first time that an RNN framework has been proposed for hyperspectral image classification. Specifically, our RNN makes use of a newly proposed activation function, parametric rectified tanh (PRetanh), for hyperspectral sequential data analysis instead of the popular tanh or rectified linear unit. The proposed activation function makes it possible to use fairly high learning rates without the risk of divergence during the training procedure. Moreover, a modified gated recurrent unit, which uses PRetanh for hidden representation, is adopted to construct the recurrent layer in our network to efficiently process hyperspectral data and reduce the total number of parameters. Experimental results on three airborne hyperspectral images suggest competitive performance in the proposed mode. In addition, the proposed network architecture opens a new window for future research, showcasing the huge potential of deep recurrent networks for hyperspectral data analysis. [ABSTRACT FROM PUBLISHER]

The deep convolutional neural network (CNN) is of great interest recently. It can provide excellent performance in hyperspectral image classification when the number of training samples is sufficiently large. In this paper, a novel pixel-pair method is proposed to significantly increase such a number, ensuring that the advantage of CNN can be actually offered. For a testing pixel, pixel-pairs, constructed by combining the center pixel and each of the surrounding pixels, are classified by the trained CNN, and the final label is then determined by a voting strategy. The proposed method utilizing deep CNN to learn pixel-pair features is expected to have more discriminative power. Experimental results based on several hyperspectral image data sets demonstrate that the proposed method can achieve better classification performance than the conventional deep learning-based method. [ABSTRACT FROM AUTHOR]

Semantic labeling (or pixel-level land-cover classification) in ultrahigh-resolution imagery (<10 cm) requires statistical models able to learn high-level concepts from spatial data, with large appearance variations. Convolutional neural networks (CNNs) achieve this goal by learning discriminatively a hierarchy of representations of increasing abstraction. In this paper, we present a CNN-based system relying on a downsample-then-upsample architecture. Specifically, it first learns a rough spatial map of high-level representations by means of convolutions and then learns to upsample them back to the original resolution by deconvolutions. By doing so, the CNN learns to densely label every pixel at the original resolution of the image. This results in many advantages, including: 1) the state-of-the-art numerical accuracy; 2) the improved geometric accuracy of predictions; and 3) high efficiency at inference time. We test the proposed system on the Vaihingen and Potsdam subdecimeter resolution data sets, involving the semantic labeling of aerial images of 9- and 5-cm resolution, respectively. These data sets are composed by many large and fully annotated tiles, allowing an unbiased evaluation of models making use of spatial information. We do so by comparing two standard CNN architectures with the proposed one: standard patch classification, prediction of local label patches by employing only convolutions, and full patch labeling by employing deconvolutions. All the systems compare favorably or outperform a state-of-the-art baseline relying on superpixels and powerful appearance descriptors. The proposed full patch labeling CNN outperforms these models by a large margin, also showing a very appealing inference time. [ABSTRACT FROM PUBLISHER]

Due to the advantages of deep learning, in this paper, a regularized deep feature extraction (FE) method is presented for hyperspectral image (HSI) classification using a convolutional neural network (CNN). The proposed approach employs several convolutional and pooling layers to extract deep features from HSIs, which are nonlinear, discriminant, and invariant. These features are useful for image classification and target detection. Furthermore, in order to address the common issue of imbalance between high dimensionality and limited availability of training samples for the classification of HSI, a few strategies such as L2 regularization and dropout are investigated to avoid overfitting in class data modeling. More importantly, we propose a 3-D CNN-based FE model with combined regularization to extract effective spectral–spatial features of hyperspectral imagery. Finally, in order to further improve the performance, a virtual sample enhanced method is proposed. The proposed approaches are carried out on three widely used hyperspectral data sets: Indian Pines, University of Pavia, and Kennedy Space Center. The obtained results reveal that the proposed models with sparse constraints provide competitive results to state-of-the-art methods. In addition, the proposed deep FE opens a new window for further research. [ABSTRACT FROM AUTHOR]

Ship detection on spaceborne images has attracted great interest in the applications of maritime security and traffic control. Optical images stand out from other remote sensing images in object detection due to their higher resolution and more visualized contents. However, most of the popular techniques for ship detection from optical spaceborne images have two shortcomings: 1) Compared with infrared and synthetic aperture radar images, their results are affected by weather conditions, like clouds and ocean waves, and 2) the higher resolution results in larger data volume, which makes processing more difficult. Most of the previous works mainly focus on solving the first problem by improving segmentation or classification with complicated algorithms. These methods face difficulty in efficiently balancing performance and complexity. In this paper, we propose a ship detection approach to solving the aforementioned two issues using wavelet coefficients extracted from JPEG2000 compressed domain combined with deep neural network (DNN) and extreme learning machine (ELM). Compressed domain is adopted for fast ship candidate extraction, DNN is exploited for high-level feature representation and classification, and ELM is used for efficient feature pooling and decision making. Extensive experiments demonstrate that, in comparison with the existing relevant state-of-the-art approaches, the proposed method requires less detection time and achieves higher detection accuracy. [ABSTRACT FROM PUBLISHER]