Deep-learning (DL) algorithms, which learn the representative and discriminative features in a hierarchical manner from the data, have recently become a hotspot in the machine-learning area and have been introduced into the geoscience and remote sensing (RS) community for RS big data analysis. Considering the low-level features (e.g., spectral and texture) as the bottom level, the output feature representation from the top level of the network can be directly fed into a subsequent classifier for pixel-based classification. As a matter of fact, by carefully addressing the practical demands in RS applications and designing the input?output levels of the whole network, we have found that DL is actually everywhere in RS data analysis: from the traditional topics of image preprocessing, pixel-based classification, and target recognition, to the recent challenging tasks of high-level semantic feature extraction and RS scene understanding.

Deep Learning for Remote Sensing Data: A Technical Tutorial on the State of the Art

Machine learning offers the potential for effective and efficient classification of remotely sensed imagery. The strengths of machine learning include the capacity to handle data of high dimensionality and to map classes with very complex characteristics. Nevertheless, implementing a machine-learning classification is not straightforward, and the literature provides conflicting advice regarding many key issues. This article therefore provides an overview of machine learning from an applied perspective. We focus on the relatively mature methods of support vector machines, single decision trees (DTs), Random Forests, boosted DTs, artificial neural networks, and k-nearest neighbours (k-NN). Issues considered include the choice of algorithm, training data requirements, user-defined parameter selection and optimization, feature space impacts and reduction, and computational costs. We illustrate these issues through applying machine-learning classification to two publically available remotely sensed dat...

https://www.tandfonline.com/doi/pdf/10.1080/01431161.2018.1433343

Implementation of machine-learning classification in remote sensing: an applied review

In this paper, we propose a spectral–spatial feature based classification (SSFC) framework that jointly uses dimension reduction and deep learning techniques for spectral and spatial feature extraction, respectively. In this framework, a balanced local discriminant embedding algorithm is proposed for spectral feature extraction from high-dimensional hyperspectral data sets. In the meantime, convolutional neural network is utilized to automatically find spatial-related features at high levels. Then, the fusion feature is extracted by stacking spectral and spatial features together. Finally, the multiple-feature-based classifier is trained for image classification. Experimental results on well-known hyperspectral data sets show that the proposed SSFC method outperforms other commonly used methods for hyperspectral image classification.

Spectral–Spatial Feature Extraction for Hyperspectral Image Classification: A Dimension Reduction and Deep Learning Approach

We propose an end-to-end framework for the dense, pixelwise classification of satellite imagery with convolutional neural networks (CNNs). In our framework, CNNs are directly trained to produce classification maps out of the input images. We first devise a fully convolutional architecture and demonstrate its relevance to the dense classification problem. We then address the issue of imperfect training data through a two-step training approach: CNNs are first initialized by using a large amount of possibly inaccurate reference data, and then refined on a small amount of accurately labeled data. To complete our framework, we design a multiscale neuron module that alleviates the common tradeoff between recognition and precise localization. A series of experiments show that our networks consider a large amount of context to provide fine-grained classification maps.

/pdf/convolutional-neural-networks-for-large-scale-remote-sensing-1y0s3rxhig.pdf

Convolutional Neural Networks for Large-Scale Remote-Sensing Image Classification

Recent research has shown that using spectral–spatial information can considerably improve the performance of hyperspectral image (HSI) classification. HSI data is typically presented in the format of 3D cubes. Thus, 3D spatial filtering naturally offers a simple and effective method for simultaneously extracting the spectral–spatial features within such images. In this paper, a 3D convolutional neural network (3D-CNN) framework is proposed for accurate HSI classification. The proposed method views the HSI cube data altogether without relying on any preprocessing or post-processing, extracting the deep spectral–spatial-combined features effectively. In addition, it requires fewer parameters than other deep learning-based methods. Thus, the model is lighter, less likely to over-fit, and easier to train. For comparison and validation, we test the proposed method along with three other deep learning-based HSI classification methods—namely, stacked autoencoder (SAE), deep brief network (DBN), and 2D-CNN-based methods—on three real-world HSI datasets captured by different sensors. Experimental results demonstrate that our 3D-CNN-based method outperforms these state-of-the-art methods and sets a new record.

/pdf/spectral-spatial-classification-of-hyperspectral-imagery-36x1sx3gn4.pdf

Spectral-spatial classification of hyperspectral imagery with 3D convolutional neural network

In this letter, a novel deep learning framework for hyperspectral image classification using both spectral and spatial features is presented. The framework is a hybrid of principal component analysis, deep convolutional neural networks (DCNNs) and logistic regression (LR). The DCNNs for hierarchically extract deep features is introduced into hyperspectral image classification for the first time. The proposed technique consists of two steps. First, feature map generation algorithm is presented to generate the spectral and spatial feature maps. Second, the DCNNs-LR classifier is trained to get useful high-level features and to fine-tune the whole model. Comparative experiments conducted over widely used hyperspectral data indicate that DCNNs-LR classifier built in this proposed deep learning framework provides better classification accuracy than previous hyperspectral classification methods.

Spectral–spatial classification of hyperspectral images using deep convolutional neural networks

The Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement mission (IMERG) has been widely evaluated. However, most of these studies focus on the ultimate merged satellite-gauge precipitation estimate and neglect the valuable intermediate estimates which directly guide the improvement of the IMERG product. This research aims to identify the error sources of the latest IMERG version 6 by evaluating the intermediate and ultimate precipitation estimates, and further examine the influences of regional topography and surface type on these errors. Results show that among six passive microwave (PMW) sensors, the Microwave Humidity Sounder (MHS) has outstanding comprehensive behavior, and Special Sensor Microwave Imager/Sounder (SSMIS) operates advanced at precipitation detection, while the Sounder for Atmospheric Profiling of Humidity in the Intertropics by Radiometry (SAPHIR) has the worst performance. More precipitation events are detected with larger quantitative uncertainty in low-lying places than in highlands, in urban and water body areas than in other places, and more in coastal areas than in inland regions. Infrared (IR) estimate has worse performance than PMW, and the precipitation detectability of IR is more sensitive to the factors of elevation and the distance to the coast, as larger critical successful index (CSI) over lowlands and coastal areas. PMW morphing and the mixing of PMW and IR algorithms partly reverse the conservative feature of the precipitation detection of PMW and IR estimates, resulting in higher probability of detection (POD) and false alert ratio (FAR). Finally, monthly gauge calibration improves most of the statistical indicators and reduces the influence of elevation and surface type factor on these errors.

Ground Validation and Error Sources Identification for GPM IMERG Product over the Southeast Coastal Regions of China

Generally, an effective ventilation system is essential to reduce coal dust disaster. However, with the implementation of carbon tax and increase of energy and operating costs, it is urgent to design a cost-effective ventilation and dust control system. In this paper, a real driving face in an excavation laneway of coal mine in China was taken as the physical model, and the temporospatial characteristics of airflow and dust dispersion is investigated for the first time to design an original conception of time-varying ventilation and dust control strategy. Specifically, computational fluid dynamic (CFD) approaches are utilized to investigate the dynamic regularity of airflow behavior and dust dispersion, and parametric studies are conducted to select the appropriate ventilation pattern, which is validated to be potential in energy saving as well as dust removal efficiency. In addition, based on the selection of key time point according to the regularity of dust concentration changed over time, the most effective type of speed function is brought out for this novel ventilation system, which reduces energy usage up to 15.11% in a ventilation period. Furthermore, the accuracy of simulation result is verified by field measurements, which demonstrates that adjusting the ventilation velocity at the appropriate time point (case 7) can effectively control the dust concentration, which performs as well as the steady flow. The research results suggest that the further understanding of temporospatial characteristics of dust dispersion is helpful for ventilation design, and significant energy savings and dust removal requirements are verified to be possible in the proposed scheme.

A Time Varying Ventilation and Dust Control Strategy Based on the Temporospatial Characteristics of Dust Dispersion

Mine safety is of primary concern in the underground coal mining system. At present, there is a lack of an efficient platform to manage the numerical simulation procedure and inherent spatiotemporal data for coal mine disasters. This necessitates the coupling of spatiotemporal model with geographic information system (GIS) in practical application. Here, a novel spatiotemporal model tightly coupled with GIS is presented to improve the model-data integration. Such tight coupling is achieved by developing a lattice Boltzmann method (LBM) based turbulent model with an underlying shared FluentEntity model within the LongRuanGIS platform. The case study and comparison with the traditional computational fluid dynamics (CFD) method demonstrated that the platform is capable and effective in providing functionalities for lattice domain decomposition, simulation, visualization and analyses, as well as improving the computational efficiency. The proposed approach and platform, promising for the disaster prevention, offer a template for future GIS-Model integration and also applicable for other underground coal mine disasters.

A Tightly Coupled GIS and Spatiotemporal Modeling for Methane Emission Simulation in the Underground Coal Mine System

With the increasing operation costs and implementation of carbon tax in the underground coal mining systems, a cost-effective ventilation system with well methane removal efficiency becomes highly required. Since the intermittent ventilation provides a novel approach in energy saving, there still exists some scientific issues. This paper adopts the design concept of frequency conversion technology (FCT) to the ventilation pattern design for the first time, and an optimized intermittent ventilation strategy is proposed. Specifically, a real excavation laneway of a coal mine in China is established as the physical model, and computational fluid dynamic (CFD) approaches are utilized to investigate the spatiotemporal characteristics of airflow behavior and methane distribution. Plus, the period of intermittency and appropriate air velocity are scientifically defined based on the conception of FCT by conducting the parametric studies. Therefore, an optimized case is brought out with well methane removal efficiency and remarkable energy reduction of 39.2% in a ventilation period. Furthermore, the simulation result is verified to be reliable by comparing with field measurements. The result demonstrates that a balance of significant energy saving and the methane removal requirement based on the concept of FTC is possible, which could in turn provide good operational support for FTC.

Hui Liu

Papers

Spectral–spatial classification of hyperspectral images using deep convolutional neural networks

Ground Validation and Error Sources Identification for GPM IMERG Product over the Southeast Coastal Regions of China

A Time Varying Ventilation and Dust Control Strategy Based on the Temporospatial Characteristics of Dust Dispersion

A Tightly Coupled GIS and Spatiotemporal Modeling for Methane Emission Simulation in the Underground Coal Mine System

A Case Study of an Optimized Intermittent Ventilation Strategy Based on CFD Modeling and the Concept of FCT