A hybrid of deep learning and hand-crafted features based approach for snow cover mapping

Nijhawan, Rahul; Das, Josodhir; Raman, Balasubramanian

doi:10.1080/01431161.2018.1519277

Cited by 44 publications

(17 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If GlacierNet output surpasses the SUS and maps DCG beyond it, then snowlines are adjusted to overlap with SUS. This adjustment is limited to the snowline area and only done to calculate evaluation indexes, such as intersection over union (IOU) values [40], recall, precision, specificity, F-measure, and accuracy [31], which are used for evaluation and defined as: where AO indicates actual output, EO indicates expected output, TP indicates true positive, FP indicates false positive, FN indicates false negative, and TN indicates true negative.…”

Section: F Evaluation Methodsmentioning

confidence: 99%

“…Deep-learning has also been used to map snow and glaciers [31], [32]. In [31], researchers used a finely-tuned AlexNet to extract features from Sentinel-2 images that were dimensionality-reduced using principal component analysis (PCA). The extracted data (i.e., glacier and background features) were then fed into a random-forest classifier.…”

Section: Deep-learning Background and Glaciernet Approachmentioning

confidence: 99%

“…The difference between these approaches is that the fully-connected layer requires massive parametrization to label each pixel compared to the locally-connected convolutional layer, as used in our study [33]. In this study, the use of our CNN segmentation model compared to previously used deep-learning [31] approaches enables us to utilize satellite and topographic data within the CNN architecture. Therefore, we do not rely on separate random forest or other classification techniques, which maybe suitable for limited geographic areas and smaller datasets (e.g., one glacier).…”

Section: Deep-learning Background and Glaciernet Approachmentioning

confidence: 99%

See 2 more Smart Citations

GlacierNet: A Deep-Learning Approach for Debris-Covered Glacier Mapping

et al. 2020

View full text Add to dashboard Cite

Rising global temperatures over the past decades is directly affecting glacier dynamics. To understand glacier fluctuations and document regional glacier-state trends, glacier-boundary detection is necessary. Debris-covered glacier (DCG) mapping, however, is notoriously difficult using conventional geospatial technology methods. Therefore, in this research for automated DCG mapping, we evaluate the utility of a convolutional neural network (CNN), which is a deep learning feed-forward neural network. The CNN inputs include Landsat satellite images, an Advanced Land Observation Satellite (ALOS) digital elevation model (DEM) and DEM-derived land-surface parameters. Our CNN based deep-learning approach named GlacierNet was designed by appropriately choosing the type, number and size of layers and filters, and encoder depth based on the properties of the input data, CNN segmentation process and empirical results. The GlacierNet was then trained using input data and corresponding glacier boundaries from the Global Land Ice Measurements from Space (GLIMS) database, and tested on glaciers in the Karakoram and Nepal Himalaya. Our results show proof-of-concept that GlacierNet reasonably identifies the boundaries of DCGs with a relatively high degree of accuracy, and that morphometric parameters improves boundary detection. INDEX TERMS Debris-covered glacier (DCG), Himalaya, Karakoram, convolutional neural network (CNN), deep-learning, image segmentation.

show abstract

Section: F Evaluation Methodsmentioning

confidence: 99%

Section: Deep-learning Background and Glaciernet Approachmentioning

confidence: 99%

Section: Deep-learning Background and Glaciernet Approachmentioning

confidence: 99%

See 1 more Smart Citation

GlacierNet: A Deep-Learning Approach for Debris-Covered Glacier Mapping

et al. 2020

View full text Add to dashboard Cite

show abstract

“…The use of hybrid approaches has also been shown to be advantageous in incorporating data from other sensors on edge nodes. Such a hybrid model where the deep learning is assisted by additional sensor sources like synthetic aperture radar (SAR) imagery and elevation like synthetic aperture radar (SAR) imagery and elevation is presented by [40]. In the context of 3D robot vision, [42] have shown that combining both linear subspace methods and deep convolutional prediction achieves improved performance along with several orders of magnitude faster runtime performance compared to the state of the art.…”

Section: Making Best Use Of Edge Computingmentioning

confidence: 99%

Deep Learning vs. Traditional Computer Vision

Mahony

Campbell

Carvalho

et al. 2019

Advances in Intelligent Systems and Computing

673

383

View full text Add to dashboard Cite

Deep Learning has pushed the limits of what was possible in the domain of Digital Image Processing. However, that is not to say that the traditional computer vision techniques which had been undergoing progressive development in years prior to the rise of DL have become obsolete. This paper will analyse the benefits and drawbacks of each approach. The aim of this paper is to promote a discussion on whether knowledge of classical computer vision techniques should be maintained. The paper will also explore how the two sides of computer vision can be combined. Several recent hybrid methodologies are reviewed which have demonstrated the ability to improve computer vision performance and to tackle problems not suited to Deep Learning. For example, combining traditional computer vision techniques with Deep Learning has been popular in emerging domains such as Panoramic Vision and 3D vision for which Deep Learning models have not yet been fully optimised.

show abstract

“…Besides, spaceborne synthetic aperture radar (SAR) data can penetrate the cloud cover, thus revealing the snow cover condition under the clouds. Based on the current method [58][59][60], total snow cover areas could be derived from the SAR observation with an accuracy of up to 98.1%. However, due to the data access policy, it would be very costly to apply for these data for long-term research at a large spatial scale.…”

Section: Challenges Of Validationmentioning

confidence: 99%

Deriving Regional Snow Line Dynamics during the Ablation Seasons 1984–2018 in European Mountains

Dietz

Kuenzer

2019

Remote Sensing

View full text Add to dashboard Cite

Snowmelt in the mid-latitude European mountains is undergoing significant spatiotemporal changes. Regional snow line elevation (RSLE) is an appropriate indicator for assessing snow cover variations in mountain areas. To derive regional snow line dynamics during the ablation seasons 1984–2018, the present study unprecedentedly introduced a readily applicable framework. The framework constitutes four steps: atmospheric and topographic correction, snow classification, RSLE retrieval, and regional snow line retreat curve (RSLRC) derivation. The developed framework has been successfully applied to 8641 satellite images acquired by Landsat, ASTER, and Sentinel-2. The results of the intra-annual regional snow line variations show that: (1) regional snow lines in the Alpine catchments preserve the longest; (2) RSLEs are lower in the northern Pyrenees than in the southern part; (3) regional snow lines persist the shortest in the Carpathian catchments; and (4) during the end of the ablation season 2018, intermediate snowfall events in the catchments Adda, Tagliamento, and Uzh are observed. In terms of the long-term inter-annual variations, significantly accelerating snow line recession is detected in the northern Pyrenean catchment Ariege. In the Alpine catchment Alpenrhein and Drac, RSLRCs are shifting towards lower accumulated air-temperature (AT) significantly, with the magnitude of −3.77 °C·a−1 (Alpenrhein) and −3.99 °C·a−1 (Drac).

show abstract

A hybrid of deep learning and hand-crafted features based approach for snow cover mapping

Cited by 44 publications

References 29 publications

GlacierNet: A Deep-Learning Approach for Debris-Covered Glacier Mapping

GlacierNet: A Deep-Learning Approach for Debris-Covered Glacier Mapping

Deep Learning vs. Traditional Computer Vision

Deriving Regional Snow Line Dynamics during the Ablation Seasons 1984–2018 in European Mountains

Contact Info

Product

Resources

About