105 research outputs found
Special section guest editorial:Advanced spectral analysis techniques and remote sensing applications
No full textSpecial section editors Sicong Liu, Francesca Bovolo, Claudio Persello, Danfeng Hong, and Alim Samat introduce the Special Section on Advanced Spectral Analysis Techniques and Remote Sensing Applications.
Spectral analysis technology represents a fundamental tool for the extraction of valuable information from spectral detection and imaging data. It can be utilized in various remote sensing applications. With the development of optical sensors, traditional spectral analysis methods may face challenges that arise due to the higher spectral resolution of multispectral to hyperspectral data, wider spectral range including ultraviolet, visible, and infrared wavelengths. Therefore, more sophisticated spectral analysis technology is urgently required. In recent decades, machine learning, especially deep learning techniques, have brought spectral analysis into the era of artificial intelligence (AI), enabling both qualitative and quantitative analysis in a more precise and robust way.
This special section aims to collect the latest developments on spectral analysis techniques and remote sensing applications. Nine excellent papers have been included in this special section, covering the following topics including: (1) spectral analysis for soil organic matter estimation; (2) spectral analysis for vegetation parameter inversion; (3) hyperspectral image super-resolution; (4) hyperspectral image classification; (5) spectral change detection.
Zhou et al. proposed an improved standard-sample calibration transfer method, in order to study the transferability of machine learning prediction models between different soil types. The results obtained by the improved model demonstrated higher accuracy of SOM prediction compared with the sample mixing method. Guo et al. used partial least squares regression (PLSR) and support vector machine regression (SVR) to establish a SOM estimation model for monitoring of soil nutrients in tensile fissures, which can provide reference for the rapid and accurate estimation of SOM in coal mining fissure zones.
Magalhães et al. compared the performance of various regression models based on Sentinel satellite images, with the aim of indirectly estimating the value of canopy water content (CWC) and equivalent water thickness (EWT) in maize more accurately. The superior performance of the AdaBoost regression (AR) model was validated in this analysis. Yasir et al. proposed a new index for estimating leaf water content based on multi-angular reflection. The effectiveness and superiority of this index were validated by 683 samples of different plant species. Jia et al. used machine learning to integrate in-situ hyperspectral data with Sentinel-2 MSI images to combine their complementary advantages, which effectively improves the accuracy of large-scale Chlorophyll-a (Chl-a) concentration estimation.
Xu et al. proposed a two-stream self-attention network (TSSA-Net) to capture global features from both multispectral and hyperspectral images. The network comprises two streams, each of which is designed to extract spatial and spectral abundance maps. The proposed method enables the generation of more effective hyperspectral image super-resolution results can be obtained.
Fang et al. proposed a novel approach, the regularized spatial-spectral transformer for domain adaptation (RSTDA), with the aim of effectively extracting spatial-spectral features from HSI data and enhancing the accuracy of cross-scene HSI classification. To this end, they adopted a smooth adversarial training strategy within the model. Atik and Atik applied explainable artificial intelligence (XAI) technology to optimal band selection in hyperspectral image classification. Their findings indicate that XAI-based methods were capable of identifying informative bands and demonstrated superior performance in subsequent tasks, as compared to other methods.
Bhattacharjee, Chakravortty, and Ghosh used the Bayesian change point detection method to identify nonlinear responses and abrupt changes in mangrove health, indicating that small environmental stresses lead to large eco-system changes over time. Results confirms that the Hurst
t
-statistics method identifies the same change points as the Bayesian approach
A Novel Context-Sensitive SVM for Classification of Remote Sensing Images
No full textIn this paper, a novel context-sensitive classification technique based on Support Vector Machines (CS-SVM) is proposed. This technique aims at exploiting the promising SVM method for classification of 2-D (or n-D) scenes by considering the spatial-context information of the pixel to be analyzed. In greater detail, the proposed architecture properly exploits the spatial-context information for: i) increasing the robustness of the learning procedure of SVMs to the noise present in the training set (mislabeled training samples); ii) regularizing the classification maps. The first property is achieved by introducing a context-sensitive term in the objective function to be minimized for defining the decision hyperplane in the SVM kernel space. The second property is obtained including in the classification procedure of a generic pattern the information of neighboring pixels. Experiments carried out on very high geometrical resolution images confirm the validity of the proposed technique
Supervised and Semi-Supervised Multi-View Canonical Correlation Analysis Ensemble for Heterogeneous Domain Adaptation in Remote Sensing Image Classification
No full textIn this paper, we present the supervised multi-view canonical correlation analysis ensemble (SMVCCAE) and its semi-supervised version (SSMVCCAE), which are novel techniques designed to address heterogeneous domain adaptation problems, i.e., situations in which the data to be processed and recognized are collected from different heterogeneous domains. Specifically, the multi-view canonical correlation analysis scheme is utilized to extract multiple correlation subspaces that are useful for joint representations for data association across domains. This scheme makes homogeneous domain adaption algorithms suitable for heterogeneous domain adaptation problems. Additionally, inspired by fusion methods such as Ensemble Learning (EL), this work proposes a weighted voting scheme based on canonical correlation coefficients to combine classification results in multiple correlation subspaces. Finally, the semi-supervised MVCCAE extends the original procedure by incorporating multiple speed-up spectral regression kernel discriminant analysis (SRKDA). To validate the performances of the proposed supervised procedure, a single-view canonical analysis (SVCCA) with the same base classifier (Random Forests) is used. Similarly, to evaluate the performance of the semi-supervised approach, a comparison is made with other techniques such as Logistic label propagation (LLP) and the Laplacian support vector machine (LapSVM). All of the approaches are tested on two real hyperspectral images, which are considered the target domain, with a classifier trained from synthetic low-dimensional multispectral images, which are considered the original source domain. The experimental results confirm that multi-view canonical correlation can overcome the limitations of SVCCA. Both of the proposed procedures outperform the ones used in the comparison with respect to not only the classification accuracy but also the computational efficiency. Moreover, this research shows that canonical correlation weighted voting (CCWV) is a valid option with respect to other ensemble schemes and that because of their ability to balance diversity and accuracy, canonical views extracted using partially joint random view generation are more effective than those obtained by exploiting disjoint random view generation
Optimizing the ground sample collection with cost-sensitive active learning for tree species classification using hyperspectral images
No full textThis study presents a cost-sensitive active learning method for optimizing the field surveys by a human expert in the classification of single tree species using hyperspectral images. The goal of the proposed method is to guide the human expert in the collection of labeled samples in order to maximize the ratio between the classification accuracy with respect to the
travelling costs. Experiments carried out in the context of a real study on forest inventory show the effectiveness of the
proposed metho
Advanced Techniques for the Classification of Very High Resolution and Hyperspectral Remote Sensing Images
No full textThis thesis is about the classification of the last generation of very high resolution (VHR) and hyperspectral remote sensing (RS) images, which are capable to acquire images characterized by very high resolution from satellite and airborne platforms. In particular, these systems can acquire VHR multispectral images characterized by a geometric resolution in the order or smaller than one meter, and hyperspectral images, characterized by hundreds of bands associated to narrow spectral channels. This type of data allows to precisely characterizing the different materials on the ground and/or the geometrical properties of the different objects (e.g., buildings, streets, agriculture fields, etc.) in the scene under investigation. This remote sensed data provide very useful information for several applications related to the monitoring of the natural environment and of human structures. However, in order to develop real-world applications with VHR and hyperspectral data, it is necessary to define automatic techniques for an efficient and effective analysis of the data. Here, we focus our attention on RS image classification, which is at the basis of most of the applications related to environmental monitoring. Image classification is devoted to translate the features that represent the information present in the data in thematic maps of the land cover types according to the solution of a pattern recognition problem. However, the huge amount of data associated with VHR and hyperspectral RS images makes the classification problem very complex and the available techniques are still inadequate to analyze these kinds of data. For this reason, the general objective of this thesis is to develop novel techniques for the analysis and the classification of VHR and hyperspectral images, in order to improve the capability to automatically extract useful information captured from these data and to exploit it in real applications. Moreover we addressed the classification of RS images in operational conditions where the available reference labeled samples are few and/or not completely reliable (which is quite common in many real problems). In particular, the following specific issues are considered in this work:
1. development of feature selection for the classification of hyperspectral images, for identifying a subset of the original features that exhibits at the same time high capability to discriminate among the considered classes and high invariance in the spatial domain of the scene;
2. classification of RS images when the available training set is not fully reliable, i.e., some labeled samples may be associated to the wrong information class (mislabeled patterns);
3. active learning techniques for interactive classification of RS images.
4. definition of a protocol for accuracy assessment in the classification of VHR images that is based on the analysis of both thematic and geometric accuracy.
For each considered topic an in deep study of the literature is carried out and the limitations of currently published methodologies are highlighted. Starting from this analysis, novel solutions are theoretically developed, implemented and applied to real RS data in order to verify their effectiveness. The obtained experimental results confirm the effectiveness of all the proposed techniques
Relevant and invariant feature selection of hyperspectral images for domain generalization
No full textThis paper presents a novel feature selection method for the analysis of hyperspectral images. The proposed method aims at selecting a subset of the original features that are both 1) relevant for the considered problem (i.e., preserve the functional relationship between input and output variables), and 2) invariant (stable) across different domains (i.e., minimize the data set shift among different domains). Domains can be associated with images collected on different areas or on the same area at different times. We propose a novel measure of domain stability, which evaluates the distance of the conditional distributions between the source and target domain. Such a measure is defined on the basis of kernel embeddings of conditional distributions and can be applied to both classification and regression problems. Experimental results show the effectiveness of the proposed method in selecting features with high generalization capabilities on the target domain
Deep Fully Convolutional Networks for the Detection of Informal Settlements in VHR Images
No full textThis 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
Active versus semi-supervised learning paradigm for the classification of remote sensing images
No full textDeep image representation learning for knowledge discovery from earth observation data archives
No full textAdvances in remote sensing (RS) technology have increased the availability of images regularly acquired by satelliteborne and airborne sensors, while free data policies support researchers to have access to massive Earth observation data archives. To automatically extract knowledge from these archives on a large-scale, deep learning (DL) based RS image representation learning (IRL) has attracted great attention. However, existing methods have limitations on: i) accurate characterization of high-level semantic content and spectral information present in RS images; ii) modelling RS image similarities by exploiting multi-label training images; iii) time efficient and scalable information extraction; iv) effective IRL under noisy training labels; and v) joint use of multiple learning tasks for describing the complex content of RS images. This thesis aims to develop advanced DL-based IRL methods to tackle these limitations, while a particular attention is devoted to image scene classification and content-based image retrieval (CBIR) problems due to their importance for large-scale knowledge discovery. In detail, we propose five DL-based IRL methods throughout the thesis. First, a multi-label classification approach is introduced to accurately describe complex spatial and spectral content of high-spatial resolution RS images, where several spectral bands are associated with varying spatial resolutions. Second, we propose an image triplet sampling method for IRL through the characterization of RS image similarities, which forms the foundation for CBIR. Among multi-label training images, this method selects a small set of the most representative and informative image triplets that lead to a decrease in computational complexity and an increase in learning speed without a significant loss in performance. Third, an approach devoted to simultaneous RS image compression and indexing is introduced for scalable CBIR. This approach characterizes hash codes of RS images on learning based compression domain; and thus prevent the requirement of decoding images prior to CBIR that can save a significant amount of time. Fourth, we propose an approach for IRL when training data includes noisy labels. By integrating generative reasoning into discriminative reasoning, our approach models the complementary characteristics of discriminative and generative reasoning, and thus prevents the interference of noisy labels during training. Fifth, a multitask learning approach is introduced to achieve IRL when multiple learning tasks are jointly utilized. Due to its loss functions and sequential optimization algorithm, this approach preserves the plasticity for each task and the stability in between learning consecutive tasks. For benchmarking the proposed methods, we introduce a large-scale multi-modal multi-label benchmark RS image archive (denoted as BigEarthNet). It includes 590,326 pairs of Sentinel-1 and Sentinel-2 image patches acquired over 10 European countries. We make BigEarthNet, its pre-trained DL models and the codes of all the methods publicly available as open source contributions of the thesis.Fortschritte in den Technologien der Fernerkundung (FK) haben zu einer erhöhten Verfügbarkeit von Bildmaterial, das von satelliten- und flugzeuggestützten Sensoren erfasst wird, geführt; gleichzeitig ermöglicht die kostenlose Freigabe von Datensätzen Forschern den Zugang zu umfangreichen Archiven mit Erdbeobachtungsdaten. Hierdurch ergibt sich ein Potential für tiefes Lernen (TF) basierte Repräsentationslernen (RL) Studien zur automatischen Wissensentdeckung aus diesen Archiven. Bestehende Methoden haben jedoch Einschränkungen in Bezug auf: i) die genaue Charakterisierung des semantischen Inhalts und der spektralen Informationen der FK-Bilder; ii) die korrekte Nutzung von FK-Bildern mit mehreren Labels während des Trainings; iii) die zeiteffiziente und skalierbare Informationsgewinnung; iv) effektives RL unter fehlerhaften Trainingslabels; und v) die kombinierte Nutzung mehrerer Lerntasks zur Beschreibung der Bildinhalte. Diese Arbeit zielt darauf ab, TF-basierte RL-Methoden zu entwickeln, um diese Defizite zu beheben, wobei ein besonderes Augenmerk auf die Klassifizierung von Bildszenen und inhaltsbasierte Bildabfragen (IB) gelegt wird. Der erste Beitrag dieser Arbeit besteht in der Entwicklung eines Multi-Label-Klassifikationsansatzes zur genauen Beschreibung des komplexen räumlichen und spektralen Inhalts hochaufgelöster FK-Bilder. Als zweiten Beitrag schlagen wir eine Bild-Tripel-Sampling-Methode für RL vor. Diese basiert auf der Charakterisierung von Bildähnlichkeiten, die grundlegend für IB sind. Unter den Trainingsbildern wählt die Methode eine kleine Anzahl verschiedener Anker sowie relevante, harte und diversifizierte Positiv- und Negativbilder aus, die zu kleineren Berechnungskomplexität ohne signifikanten Performanceverlust führen. Im dritten Beitrag wird ein Ansatz zur gleichzeitigen FK-Bildkompression und Indizierung für skalierbares IB vorgestellt. Unser Ansatz charakterisiert Hash-Codes von FK-Bildern auf einer lernbasierten Kompressionsdomäne und erspart somit die Dekodierung von Bildern vor der IB, was zu einer erheblichen Zeitersparnis führen kann. Als vierten Beitrag schlagen wir einen Ansatz für RL vor, für den Fall, dass die Trainingsdaten fehlerhafte Labels enthalten. Durch die Kombination von generativen und diskriminativen Modellierungen nutzt unser Ansatz ihre komplementären Eigenschaften, um die Störung durch fehlerhafte Labels während des Trainings zu verhindern. Im fünften Beitrag wird ein Multitask-Lernansatz eingeführt, bei dem mehrere Lerntasks kombiniert verwendet werden. Aufgrund seiner Verlustfunktionen und seines sequentiellen Optimierungsalgorithmus bewahrt dieser Ansatz die Plastizität für jeden einzelnen Lerntask und die Stabilität zwischen aufeinanderfolgenden Lerntasks. Für das Benchmarking der vorgeschlagenen Methoden besteht der letzte Beitrag dieser Arbeit in der Erstellung von BigEarthNet, dem ersten groß angelegten multimodalen Multi-Label-Benchmark-Archiv in FK. Wir stellen BigEarthNet, seine vortrainierten TF-Modelle und die Codes aller Methoden als Open-Source-Beiträge der Dissertation öffentlich zur Verfügung.EC/H2020/759764/EU/Accurate and Scalable Processing of Big Data in Earth Observation/BigEart
- …
