Unsupervised Machine Learning for Improved Delaunay Triangulation

Physical oceanography models rely heavily on grid discretization. It is known that unstructured grids perform well in dealing with boundary fitting problems in complex nearshore regions. However, it is time-consuming to find a set of unstructured grids in specific ocean areas, particularly in the case of land areas that are frequently changed by human construction. In this work, an attempt was made to use machine learning for the optimization of the unstructured triangular meshes formed with Delaunay triangulation in the global ocean field, so that the triangles in the triangular mesh were closer to equilateral triangles, the long, narrow triangles in the triangular mesh were reduced, and the mesh quality was improved. Specifically, we used Delaunay triangulation to generate the unstructured grid, and then developed a K-means clustering-based algorithm to optimize the unstructured grid. With the proposed method, unstructured meshes were generated and optimized for global oceans, small sea areas, and the South China Sea estuary to carry out data experiments. The results suggested that the proportion of triangles with a triangle shape factor greater than 0.7 amounted to 77.80%, 79.78%, and 79.78%, respectively, in the unstructured mesh. Meanwhile, the proportion of long, narrow triangles in the unstructured mesh was decreased to 8.99%, 3.46%, and 4.12%, respectively.

Download Full-text

TempestExtremes v1.0: A Framework for Scale-Insensitive Pointwise Feature Tracking on Unstructured Grids

10.5194/gmd-2016-217 ◽

2016 ◽

Cited By ~ 4

Author(s):

Paul A. Ullrich ◽

Colin M. Zarzycki

Keyword(s):

Open Source Software ◽

Feature Tracking ◽

Unstructured Grid ◽

Unstructured Grids ◽

Software Framework ◽

Extratropical Cyclones ◽

Easterly Waves ◽

The Past ◽

Detection Schemes ◽

Core Functionality

Abstract. This paper describes a new open-source software framework for automated pointwise feature tracking that is applicable to a wide array of climate datasets using either structured or unstructured grids. Common climatological pointwise features include tropical cyclones, extratropical cyclones and tropical easterly waves. To enable support for a wide array of detection schemes, a suite of algorithmic kernels have been developed that capture the core functionality of algorithmic tracking routines from throughout literature. A review of efforts related to pointwise feature tracking from the past three decades is included. Selected results using both reanalysis datasets and unstructured grid simulations are provided.

Download Full-text

A Combined Global Coarsening Method for 3D Multigrid Applications

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.236-237.1049 ◽

2012 ◽

Vol 236-237 ◽

pp. 1049-1053

Author(s):

Zong Zhe Li ◽

Zheng Hua Wang ◽

Lu Yao ◽

Wei Cao

Keyword(s):

Aspect Ratio ◽

Present Method ◽

Unstructured Grid ◽

Unstructured Grids ◽

Coarse Grid ◽

3D Flow ◽

High Quality ◽

Coarse Grids

An automatic agglomeration methodology to generate coarse grids for 3D flow solutions on anisotropic unstructured grids has been introduced in this paper. The algorithm combines isotropic octree based coarsening and anisotropic directional agglomeration to yield a desired coarsening ratio and high quality of coarse grids, which developed for cell-centered multigrid applications. This coarsening strategy developed is presented on an unstructured grid over 3D ONERA M6 wing. It is shown that the present method provides suitable coarsening ratio and well defined aspect ratio cells at all coarse grid levels.

Download Full-text

A New Agglomeration Method for Isotropic Unstructured Grids

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.241-244.2957 ◽

2012 ◽

Vol 241-244 ◽

pp. 2957-2961

Author(s):

Zong Zhe Li ◽

Zheng Hua Wang ◽

Wei Cao ◽

Lu Yao

Keyword(s):

Aspect Ratio ◽

Euler Equations ◽

Multigrid Method ◽

Unstructured Grid ◽

Convergence Rates ◽

Unstructured Grids ◽

Common Vertex ◽

Finite Volume Scheme ◽

High Quality ◽

Coarse Grids

A robust aspect ratio based agglomeration algorithm to generate high quality coarse grids for unstructured grid is proposed in this paper. The algorithm focuses on multigrid techniques for the numerical solution of Euler equations, which conform to cell-centered finite volume scheme, combines isotropic vertex-based agglomeration to yield large increases in convergence rates. Aspect ratio is used as fusing weight to capture the degree of cell convexity and give an indication of cell quality, agglomerating isotropic cells sharing a common vertex. Consequently, we conduct agglomeration multigrid method to solve Euler equations on 2D isotropic unstructured grid, and compare the results with MGridGen

Download Full-text

Geometric Accuracy Prediction for Additive Manufacturing Through Machine Learning of Triangular Mesh Data

Volume 1: Additive Manufacturing; Manufacturing Equipment and Systems; Bio and Sustainable Manufacturing ◽

10.1115/msec2019-3050 ◽

2019 ◽

Cited By ~ 4

Author(s):

Nathan Decker ◽

Qiang Huang

Keyword(s):

Machine Learning ◽

Additive Manufacturing ◽

Three Dimensional ◽

Triangular Mesh ◽

Geometric Accuracy ◽

New Approach ◽

Machine Learning Model ◽

3D Printed ◽

3D Shapes ◽

Complex Parts

Abstract While additive manufacturing has seen tremendous growth in recent years, a number of challenges remain, including the presence of substantial geometric differences between a three dimensional (3D) printed part, and the shape that was intended. There are a number of approaches for addressing this issue, including statistical models that seek to account for errors caused by the geometry of the object being printed. Currently, these models are largely unable to account for errors generated in freeform 3D shapes. This paper proposes a new approach using machine learning with a set of predictors based on the geometric properties of the triangular mesh file used for printing. A direct advantage of this method is the simplicity with which it can describe important properties of a 3D shape and allow for predictive modeling of dimensional inaccuracies for complex parts. To evaluate the efficacy of this approach, a sample dataset of 3D printed objects and their corresponding deviations was generated. This dataset was used to train a random forest machine learning model and generate predictions of deviation for a new object. These predicted deviations were found to compare favorably to the actual deviations, demonstrating the potential of this approach for applications in error prediction and compensation.

Download Full-text

TempestExtremes: a framework for scale-insensitive pointwise feature tracking on unstructured grids

Geoscientific Model Development ◽

10.5194/gmd-10-1069-2017 ◽

2017 ◽

Vol 10 (3) ◽

pp. 1069-1090 ◽

Cited By ~ 39

Author(s):

Paul A. Ullrich ◽

Colin M. Zarzycki

Keyword(s):

Open Source Software ◽

Feature Tracking ◽

Unstructured Grid ◽

Unstructured Grids ◽

Software Framework ◽

Extratropical Cyclones ◽

Easterly Waves ◽

The Past ◽

Detection Schemes ◽

Core Functionality

Abstract. This paper describes a new open-source software framework for automated pointwise feature tracking that is applicable to a wide array of climate datasets using either structured or unstructured grids. Common climatological pointwise features include tropical cyclones, extratropical cyclones and tropical easterly waves. To enable support for a wide array of detection schemes, a suite of algorithmic kernels have been developed that capture the core functionality of algorithmic tracking routines throughout the literature. A review of efforts related to pointwise feature tracking from the past 3 decades is included. Selected results using both reanalysis datasets and unstructured grid simulations are provided.

Download Full-text

Sizing Function Based on Machine Learning for Unstructured Mesh Generation

Proceedings of the 5th World Congress on Mechanical, Chemical, and Material Engineering ◽

10.11159/icmie19.117 ◽

2019 ◽

Author(s):

Xiang Gao ◽

Chuanfu Xu

Keyword(s):

Machine Learning ◽

Mesh Generation ◽

Unstructured Mesh ◽

Unstructured Mesh Generation

Download Full-text

Deep-sea sediments of the global ocean

10.5194/essd-2020-22 ◽

2020 ◽

Cited By ~ 1

Author(s):

Markus Diesing

Keyword(s):

Machine Learning ◽

Spatial Distribution ◽

Deep Sea ◽

Learning Algorithm ◽

Marine Species ◽

Global Ocean ◽

Teaching Management ◽

Sediment Distribution ◽

Deep Sea Sediments ◽

Seafloor Sediment

Abstract. Although the deep-sea floor accounts for more than 70 % of the Earth's surface, there has been little progress in relation to deriving maps of seafloor sediment distribution based on transparent, repeatable and automated methods such as machine learning. A new digital map of the spatial distribution of seafloor lithologies in the deep sea below 500 m water depth is presented to address this shortcoming. The lithology map is accompanied by estimates of the probability of the most probable class, which may be interpreted as a spatially-explicit measure of confidence in the predictions, and probabilities for the occurrence of seven lithology classes (Calcareous sediment, Clay, Diatom ooze, Lithogenous sediment, Mixed calcareous-siliceous ooze, Radiolarian ooze and Siliceous mud). These map products were derived by the application of the Random Forest machine learning algorithm to a homogenised dataset of seafloor lithology samples and global environmental predictor variables that were selected based on the current understanding of the controls on the spatial distribution of deep-sea sediments. The overall accuracy of the lithology map is 69.5 %, with 95 % confidence limits of 67.9 % and 71.1 %. It is expected that the map products are useful for various purposes including, but not limited to, teaching, management, spatial planning, design of marine protected areas and as input for global spatial predictions of marine species distributions and seafloor sediment properties. The map products are available at https://doi.org/10.1594/PANGAEA.911692 (Diesing, 2020).

Download Full-text

Facial Expressions Recognition Based on Delaunay Triangulation of Landmark and Machine Learning

Traitement du signal ◽

10.18280/ts.380602 ◽

2021 ◽

Vol 38 (6) ◽

pp. 1575-1586

Author(s):

Farid Ayeche ◽

Adel Alti

Keyword(s):

Machine Learning ◽

Facial Expressions ◽

Delaunay Triangulation ◽

Connected Graph ◽

Recognition Rate ◽

Machine Learning Techniques ◽

Expression Recognition ◽

Technological Advances ◽

Learning Techniques ◽

Machine Learning Approach

Facial expressions can tell a lot about an individual’s emotional state. Recent technological advances opening avenues for automatic Facial Expression Recognition (FER) based on machine learning techniques. Many works have been done on FER for the classification of facial expressions. However, the applicability to more naturalistic facial expressions remains unclear. This paper intends to develop a machine learning approach based on the Delaunay triangulation to extract the relevant facial features allowing classifying facial expressions. Initially, from the given facial image, a set of discriminative landmarks are extracted. Along with this, a minimal landmark connected graph is also extracted. Thereby, from the connected graph, the expression is represented by a one-dimensional feature vector. Finally, the obtained vector is subject for classification by six well-known classifiers (KNN, NB, DT, QDA, RF and SVM). The experiments are conducted on four standard databases (CK+, KDEF, JAFFE and MUG) to evaluate the performance of the proposed approach and find out which classifier is better suited to our system. The QDA approach based on the Delaunay triangulation has a high accuracy of 96.94% since it only supports non-zero pixels, which increases the recognition rate.

Download Full-text

Deep Learning for the derivation of GNSS Reflectometry global ocean wind speed

10.5194/egusphere-egu21-4665 ◽

2021 ◽

Author(s):

Milad Asgarimehr ◽

Caroline Arnold ◽

Felix Stiehler ◽

Tobias Weigel ◽

Chris Ruf ◽

...

Keyword(s):

Machine Learning ◽

Remote Sensing ◽

Deep Learning ◽

Wind Speed ◽

Sampling Rate ◽

Machine Learning Techniques ◽

Global Ocean ◽

Deep Model ◽

Learning Techniques ◽

Unseen Data

The Global Navigation Satellite System Reflectometry (GNSS-R) is a novel remote sensing technique exploiting GNSS signals after reflection off the Earth's surface. The capability of spaceborne GNSS-R to monitor ocean state and the surface wind is recently well demonstrated, which offers an unprecedented sampling rate and much robustness during rainfall. The Cyclone GNSS (CyGNSS) is the first spaceborne mission fully dedicated to GNSS-R, launched in December 2016.Thanks to the low development costs of the GNSS-R satellite missions as well as the capability of tracking multiple reflected signals from numerous GNSS transmitters, the GNSS-R datasets are much bigger compared to those from conventional remote sensing techniques. The CyGNSS provides a high number of unique samples in the order of a few millions monthly.&#160; Deep learning can therefore be implemented in GNSS-R even more efficiently than other remote sensing domains. With the upcoming GNSS-R CubeSats, the data volume is expected to increase in the near future and GNSS-R &#8220;Big data&#8221; can be a future challenge. Deep learning methods are additionally able to correct the potential effects, both technical and geophysical, dictated by data empirically when the mechanisms are not well described by the theoretical knowledge. This poses the question if GNSS-R should embrace deep learning and can benefit from this modern data scientific method like other Earth Observation domains.The receivers onboard CyGNSS cross-correlate the reflected signals received at a nadir antenna to a locally generated replica. The cross-correlation power at a range of the signal delay and Doppler frequency shift is the observational output of the receivers being called delay-Doppler Maps (DDMs). The mapped power is inversely proportional to the ocean roughness and consequently surface winds.Few recent studies innovatively show some merits of machine learning techniques for the derivations of ocean winds from the DDMs. However, the capability of machine learning techniques, especially deep learning for an operational data derivation needs to be better characterized. Normally, the operational retrieval algorithms are developed based on an existing dataset and are supposed to operate on the upcoming measurements. Therefore, machine learning-based models are supposed to generalize well on the unseen data in future periods. Herein, we aim at the characterization of deep learning capabilities for these GNSS-R operational purposes.In this interdisciplinary study, we present a deep learning algorithm processing the CyGNSS measurements to derive wind speed data. The model is supposed to meet an acceptable level of generalization on the upcoming unseen data, and alternatively can be used as an operational processing algorithm. We propose a deep model based on convolutional and fully connected layers processing the DDMs besides ancillary input features. The model leads to the so-far best quality of global wind speed estimates using GNSS-R measurements with a general root mean square error of 1.3 m/s over unseen data in a time span different from that of the training data.

Download Full-text

Application of machine learning to map the global distribution of deep-sea sediments

10.5194/egusphere-egu2020-6755 ◽

2020 ◽

Author(s):

Markus Diesing

Keyword(s):

Machine Learning ◽

Deep Sea ◽

Learning Algorithm ◽

Predictor Variable ◽

Global Ocean ◽

Predictor Variables ◽

Maximum Temperature ◽

Sea Floor ◽

Deep Sea Sediments ◽

Calcareous Sediment

The deep-sea floor accounts for >90% of seafloor area and >70% of the Earth&#8217;s surface. It acts as a receptor of the particle flux from the surface layers of the global ocean, is a place of biogeochemical cycling, records environmental and climate conditions through time and provides habitat for benthic organisms. Maps of the spatial patterns of deep-sea sediments are therefore a major prerequisite for many studies addressing aspects of deep-sea sedimentation, biogeochemistry, ecology and related fields.A new digital map of deep-sea sediments of the global ocean is presented. The map was derived by applying the Random Forest machine-learning algorithm to published sample data of seafloor lithologies and environmental predictor variables. The selection of environmental predictors was initially based on the current understanding of the controls on the distribution of deep-sea sediments and the availability of data. A predictor variable selection process ensured that only important and uncorrelated variables were employed in the model. The three most important predictor variables were sea-surface maximum salinity, sea-floor maximum temperature and bathymetry. The occurrence probabilities of seven seafloor lithologies (Calcareous sediment, Clay, Diatom ooze, Lithogenous sediment, Mixed calcareous-siliceous ooze, Radiolarian ooze and Siliceous mud) were spatially predicted. The final map shows the most probable seafloor lithology and an associated probability value, which may be viewed as a spatially explicit measure of map confidence. An assessment of the accuracy of the map was based on a test set of observations not used for model training. Overall map accuracy was 69.5% (95% confidence interval: 67.9% - 71.1%). The sea-floor lithology map bears some resemblance with previously published hand-drawn maps in that the distribution of Calcareous sediment, Clay and Diatom ooze are very similar. Clear differences were however also noted: Most strikingly, the map presented here does not display a band of Radiolarian ooze in the equatorial Pacific.The probability surfaces of individual seafloor lithologies, the categorical map of the seven mapped lithologies and the associated map confidence will be made freely available. It is hoped that they form a useful basis for research pertaining to deep-sea sediments.

Download Full-text