scholarly journals Unsupervised Deep Learning for Landslide Detection from Multispectral Sentinel-2 Imagery

2021 ◽  
Vol 13 (22) ◽  
pp. 4698
Author(s):  
Hejar Shahabi ◽  
Maryam Rahimzad ◽  
Sepideh Tavakkoli Piralilou ◽  
Omid Ghorbanzadeh ◽  
Saied Homayouni ◽  
...  
Keyword(s):  
Deep Learning ◽  
Training Data ◽  
New Approach ◽  
Digital Elevation ◽  
Unsupervised Deep Learning ◽  
Elevation Model ◽  
High Level ◽  
The Impact ◽  
Noise Fraction ◽  
Sentinel 2

This paper proposes a new approach based on an unsupervised deep learning (DL) model for landslide detection. Recently, supervised DL models using convolutional neural networks (CNN) have been widely studied for landslide detection. Even though these models provide robust performance and reliable results, they depend highly on a large labeled dataset for their training step. As an alternative, in this paper, we developed an unsupervised learning model by employing a convolutional auto-encoder (CAE) to deal with the problem of limited labeled data for training. The CAE was used to learn and extract the abstract and high-level features without using training data. To assess the performance of the proposed approach, we used Sentinel-2 imagery and a digital elevation model (DEM) to map landslides in three different case studies in India, China, and Taiwan. Using minimum noise fraction (MNF) transformation, we reduced the multispectral dimension to three features containing more than 80% of scene information. Next, these features were stacked with slope data and NDVI as inputs to the CAE model. The Huber reconstruction loss was used to evaluate the inputs. We achieved reconstruction losses ranging from 0.10 to 0.147 for the MNF features, slope, and NDVI stack for all three study areas. The mini-batch K-means clustering method was used to cluster the features into two to five classes. To evaluate the impact of deep features on landslide detection, we first clustered a stack of MNF features, slope, and NDVI, then the same ones plus with the deep features. For all cases, clustering based on deep features provided the highest precision, recall, F1-score, and mean intersection over the union in landslide detection.

scholarly journals Potential of Sentinel-2 Images for Estimating of Soil Resistivity over Agricultural Fields

Proceedings ◽  
2019 ◽  
Vol 24 (1) ◽  
pp. 18
Author(s):  
Remy Fieuzal ◽  
Vincent Bustillo ◽  
David Collado ◽  
Gerard Dedieu
Keyword(s):  
Electrical Resistivity ◽  
Near Infrared ◽  
Soil Layers ◽  
Digital Elevation ◽  
Combined Use ◽  
Input Variables ◽  
Elevation Model ◽  
High Level ◽  
Sentinel 2 ◽  
Soil Electrical Resistivity

The aim of this study is to assess the possibilities of the VNIR (Visible and Near InfraRed) and SWIR (Short Wavelength InfraRed) satellite data for estimating intra-plot patterns of soil electrical resistivity consistent with ground measurements. The methodology is based on optical reflectances that constitute the input variables of random forest, alone or in combination with parameters derived from a digital elevation model (DEM). Over a field located in southwestern France, the results show high level of accuracy for the 0–50 and 0–100 cm soil layers (with R² of 0.69 and 0.59, and a relative RMSE of 18% and 16%, respectively), the performances being lower for the 0–170 cm layer (R² of 0.39, relative RMSE of 20%). The combined use of optical reflectances with parameters derived from the DEM slightly improves the performances, whatever the considered layer. The influence of each reflectance on soil electrical resistivity estimates is finally analyzed, showing that the wavelengths acquired in the SWIR have a relative higher importance than VNIR reflectance.

scholarly journals Copy-Move Forgery Detection (CMFD) Using Deep Learning for Image and Video Forensics

2021 ◽  
Vol 7 (3) ◽  
pp. 59
Author(s):  
Yohanna Rodriguez-Ortega ◽  
Dora M. Ballesteros ◽  
Diego Renza
Keyword(s):  
Deep Learning ◽  
Transfer Learning ◽  
Training Data ◽  
Video Editing ◽  
Forgery Detection ◽  
Copy Move Forgery Detection ◽  
The Impact ◽  
And Training ◽  
Selection Of ◽  
Traditional Image

With the exponential growth of high-quality fake images in social networks and media, it is necessary to develop recognition algorithms for this type of content. One of the most common types of image and video editing consists of duplicating areas of the image, known as the copy-move technique. Traditional image processing approaches manually look for patterns related to the duplicated content, limiting their use in mass data classification. In contrast, approaches based on deep learning have shown better performance and promising results, but they present generalization problems with a high dependence on training data and the need for appropriate selection of hyperparameters. To overcome this, we propose two approaches that use deep learning, a model by a custom architecture and a model by transfer learning. In each case, the impact of the depth of the network is analyzed in terms of precision (P), recall (R) and F1 score. Additionally, the problem of generalization is addressed with images from eight different open access datasets. Finally, the models are compared in terms of evaluation metrics, and training and inference times. The model by transfer learning of VGG-16 achieves metrics about 10% higher than the model by a custom architecture, however, it requires approximately twice as much inference time as the latter.

scholarly journals An algorithm for delineating and extracting hillslopes and hillslope width functions from gridded elevation data

2011 ◽  
Vol 8 (5) ◽  
pp. 8865-8901
Author(s):  
P. Noel ◽  
A. N. Rousseau ◽  
C. Paniconi
Keyword(s):  
Three Dimensional ◽  
Rainfall Runoff ◽  
Dimensional Representation ◽  
New Approach ◽  
One Dimensional ◽  
Digital Elevation ◽  
Elevation Data ◽  
Average Profile ◽  
Elevation Model ◽  
Digital Elevation Model Resolution

Abstract. Subdivision of catchment into appropriate hydrological units is essential to represent rainfall-runoff processes in hydrological modelling. The commonest units used for this purpose are hillslopes (e.g. Fan and Bras, 1998; Troch et al., 2003). Hillslope width functions can therefore be utilised as one-dimensional representation of three-dimensional landscapes by introducing profile curvatures and plan shapes. An algorithm was developed to delineate and extract hillslopes and hillslope width functions by introducing a new approach to calculate an average profile curvature and plan shape. This allows the algorithm to be independent of digital elevation model resolution and to associate hillslopes to nine elementary landscapes according to Dikau (1989). This algortihm was tested on two flat and steep catchments of the province of Quebec, Canada. Results showed great area coverage for hillslope width function over individual hillslopes and entire watershed.

scholarly journals A New Approach to the Quality Control of Slope and Aspect Classes Derived from Digital Elevation Models

2021 ◽  
Vol 13 (11) ◽  
pp. 2069
Author(s):  
M. V. Alba-Fernández ◽  
F. J. Ariza-López ◽  
M. D. Jiménez-Gamero
Keyword(s):  
Quality Control ◽  
Hypothesis Test ◽  
Simulated Data ◽  
Slope Aspect ◽  
New Approach ◽  
Reference Set ◽  
Digital Elevation ◽  
Using Data ◽  
Elevation Model ◽  
Degree Of Similarity

The usefulness of the parameters (e.g., slope, aspect) derived from a Digital Elevation Model (DEM) is limited by its accuracy. In this paper, a thematic-like quality control (class-based) of aspect and slope classes is proposed. A product can be compared against a reference dataset, which provides the quality requirements to be achieved, by comparing the product proportions of each class with those of the reference set. If a distance between the product proportions and the reference proportions is smaller than a small enough positive tolerance, which is fixed by the user, it will be considered that the degree of similarity between the product and the reference set is acceptable, and hence that its quality meets the requirements. A formal statistical procedure, based on a hypothesis test, is developed and its performance is analyzed using simulated data. It uses the Hellinger distance between the proportions. The application to the slope and aspect is illustrated using data derived from a 2×2 m DEM (reference) and 5×5 m DEM in Allo (province of Navarra, Spain).

scholarly journals Unsupervised Deep Learning via Affinity Diffusion

2020 ◽  
Vol 34 (07) ◽  
pp. 11029-11036
Author(s):  
Jiabo Huang ◽  
Qi Dong ◽  
Shaogang Gong ◽  
Xiatian Zhu
Keyword(s):  
Deep Learning ◽  
State Of The Art ◽  
General Purpose ◽  
Training Data ◽  
Learning Approach ◽  
Model Learning ◽  
Feature Representations ◽  
Discriminative Feature ◽  
Training Samples ◽  
Unsupervised Deep Learning

Convolutional neural networks (CNNs) have achieved unprecedented success in a variety of computer vision tasks. However, they usually rely on supervised model learning with the need for massive labelled training data, limiting dramatically their usability and deployability in real-world scenarios without any labelling budget. In this work, we introduce a general-purpose unsupervised deep learning approach to deriving discriminative feature representations. It is based on self-discovering semantically consistent groups of unlabelled training samples with the same class concepts through a progressive affinity diffusion process. Extensive experiments on object image classification and clustering show the performance superiority of the proposed method over the state-of-the-art unsupervised learning models using six common image recognition benchmarks including MNIST, SVHN, STL10, CIFAR10, CIFAR100 and ImageNet.

scholarly journals Effect of DEM Interpolation Neighbourhood on Terrain Factors

2019 ◽  
Vol 8 (1) ◽  
pp. 30 ◽  
Author(s):  
Ying Zhu ◽  
Xuejun Liu ◽  
Jing Zhao ◽  
Jianjun Cao ◽  
Xiaolei Wang ◽  
...  
Keyword(s):  
Interpolation Method ◽  
Simulated Data ◽  
Topographic Features ◽  
Structure And Morphology ◽  
Digital Elevation ◽  
Different Types ◽  
Slope Error ◽  
Elevation Model ◽  
The Impact ◽  
Mls Method

Topographic factors such as slope and aspect are essential parameters in depicting the structure and morphology of a terrain surface. We study the effect of the number of points in the neighbourhood of a digital elevation model (DEM) interpolation method on mean slope, mean aspect, and RMSEs of slope and aspect from the interpolated DEM. As the moving least squares (MLS) method can maintain the inherent properties and other characteristics of a surface, this method is chosen for DEM interpolation. Three areas containing different types of topographic features are selected for study. Simulated data from a Gauss surface is also used for comparison. First, the impact of the number of points on the DEM root mean square error (RMSE) is analysed. The DEM RMSE in the three study areas decreases gradually with the number of points in the neighbourhood. In addition, the effect of the number of points in the neighbourhood on mean slope and mean aspect was studied across varying topographies through regression analysis. The two variables respond differently to changes in terrain. However, the RMSEs of the slope and aspect in all study areas are logarithmically related to the number of points in the neighbourhood and the values decrease uniformly as the number of points in the neighbourhood increases. With more points in the neighbourhood, the RMSEs of the slope and aspect are not sensitive to topography differences and the same trends are observed for the three studied quantities. Results for the Gauss surface are similar. Finally, this study analyses the spatial distribution of slope and aspect errors. The slope error is concentrated in ridges, valleys, steep-slope areas, and ditch edges while the aspect error is concentrated in ridges, valleys, and flat regions. With more points in the neighbourhood, the number of grid cells in which the slope error is greater than 15° is gradually reduced. With similar terrain types and data sources, if the calculation efficiency is not a concern, sufficient points in the spatial autocorrelation range should be analysed in the neighbourhood to maximize the accuracy of the slope and aspect. However, selecting between 10 and 12 points in the neighbourhood is economical.

scholarly journals Impact of Spatial Resolution of Digital Elevation Model on Landslide Susceptibility Mapping: A case Study in Kullu Valley, Himalayas

Geosciences ◽  
2019 ◽  
Vol 9 (8) ◽  
pp. 360 ◽  
Author(s):  
Sansar Raj ◽  
Thimmaiah
Keyword(s):  
Digital Elevation Model ◽  
Spatial Resolution ◽  
Landslide Susceptibility ◽  
Susceptibility Mapping ◽  
Landslide Susceptibility Mapping ◽  
Mountainous Regions ◽  
Digital Elevation ◽  
Kullu Valley ◽  
Elevation Model ◽  
The Impact

Landslides are one of the most damaging geological hazards in mountainous regions such as the Himalayas. The Himalayan region is, tectonically, the most active region in the world that is highly vulnerable to landslides and associated hazards. Landslide susceptibility mapping (LSM) is a useful tool for understanding the probability of the spatial distribution of future landslide regions. In this research, the landslide inventory datasets were collected during the field study of the Kullu valley in July 2018, and 149 landslide locations were collected as global positioning system (GPS) points. The present study evaluates the LSM using three different spatial resolution of the digital elevation model (DEM) derived from three different sources. The data-driven traditional frequency ratio (FR) model was used for this study. The FR model was used for this research to assess the impact of the different spatial resolution of DEMs on the LSM. DEM data was derived from Advanced Land Observing Satellite-1 (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) ALOS-PALSAR for 12.5 m, the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global for 30 m, and the Shuttle Radar Topography Mission (SRTM) for 90 m. As an input, we used eight landslide conditioning factors based on the study area and topographic features of the Kullu valley in the Himalayas. The ASTER-Global 30m DEM showed higher accuracy of 0.910 compared to 0.839 for 12.5 m and 0.824 for 90 m DEM resolution. This study shows that that 30 m resolution is better suited for LSM for the Kullu valley region in the Himalayas. The LSM can be used for mitigation and future planning for spatial planners and developmental authorities in the region.

scholarly journals Assessment of Changes in a Viewshed in the Western Carpathians Landscape as a Result of Reforestation

Land ◽  
2020 ◽  
Vol 9 (11) ◽  
pp. 430
Author(s):  
Michał Sobala ◽  
Urszula Myga-Piątek ◽  
Bartłomiej Szypuła
Keyword(s):  
Digital Elevation Model ◽  
Forest Succession ◽  
Western Carpathians ◽  
Surface Model ◽  
Digital Surface Model ◽  
Mountainous Areas ◽  
Viewshed Analysis ◽  
Digital Elevation ◽  
Elevation Model ◽  
The Impact

A viewshed analysis is of great importance in mountainous areas characterized by high landscape values. The aim of this research was to determine the impact of reforestation occurring on former pasturelands on changes in the viewshed, and to quantify changes in the surface of glades. We combine a horizontal and a vertical approach to landscape analysis. The changes in non-forest areas and the viewshed from viewpoints located in glades were calculated using historical cartographic materials and a more recent Digital Elevation Model and Digital Surface Model. An analysis was conducted using a Visibility tool in ArcGIS. The non-forest areas decreased in the period 1848–2015. The viewshed in the majority of viewpoints also decreased in the period 1848–2015. In the majority of cases, the maximal viewsheds were calculated in 1879/1885 and 1933 (43.8% of the analyzed cases), whereas the minimal ones were calculated in 2015 (almost 57.5% of analyzed cases). Changes in the viewshed range from 0.2 to 23.5 km2 with half the cases analyzed being no more than 1.4 km2. The results indicate that forest succession on abandoned glades does not always cause a decline in the viewshed. Deforestation in neighboring areas may be another factor that has an influence on the decline.

Sign in / Sign up

Export Citation Format

Share Document