Analyse de l'incertitude et de la précision thématique de classifications GEOBIA d'une image WorldView-2

2018 ◽  
pp. 19-37
Author(s):  
François Messner ◽  
Jeannine Corbonnois ◽  
Fanny Stella Tchitouo Ntenzou
Keyword(s):  
Random Forest ◽  
A Posteriori ◽  
Distribution Spatiale ◽  
Probabilité A Posteriori

L'évaluation de la précision des cartes thématiques produites par télédétection est une finalité de tout processus de classification modélisant le paysage. Reposant traditionnellement sur la matrice de confusion, elle peut être complétée par des méthodes alternatives plus à même de prendre en compte le biais relatif à la sélection des échantillons d'apprentissage, ainsi que par l'emploi d'approches représentant spatialement l'incertitude inhérente aux classifications. Une telle démarche est adoptée dans cet article, en évaluant la précision à l'aide des estimateurs du Maximum de Probabilité a Posteriori, puis en déterminant, pour chaque unité de carte, des mesures d'incertitude : l'entropie quadratique, la divergence de Kullback-Leibler et un indice de concordance qualitatif. Ces traitements sont analysés et comparés selon 3 classifieurs, Random Forest, C5.0 et l'Analyse Discriminante Linéaire et selon 4 stratégies de classification : classifieurs seuls, classifieurs avec procédure de bagging, classifieurs avec procédure d'apprentissage actif et classifieurs avec procédure d'apprentissage actif et de bagging. Les résultats obtenus soulignent la complémentarité des estimateurs de précision pour mettre en évidence un biais dans l'évaluation de la précision ou dans la détermination des probabilités a posteriori, et justifie la prise en considération des indices d'incertitude comme source d'informations sur la distribution spatiale des erreurs de cartographie.

La susceptibilité aux glissements de terrain dans la ville d’Al Hoceima et sa périphérie : application la méthode de la théorie de l’évidence

GEOMATICA ◽  
2020 ◽  
Author(s):  
Taoufik BYOU ◽  
Khalid OBDA ◽  
Ali TAOUS ◽  
Ilias OBDA
Keyword(s):  
Receiver Operating Characteristic ◽  
Operating Characteristic ◽  
A Priori ◽  
A Posteriori ◽  
Mots Clés ◽  
Glissement De Terrain ◽  
Glissements De Terrain ◽  
Al Hoceima ◽  
Receiver Operating ◽  
Probabilité A Posteriori

Résumé : Le Rif Marocain en général et la ville d’Al Hoceima et sa périphérie urbaine plus particulièrement, connaissent fréquemment des aléas géomorphologiques, notamment les glissements de terrain qui entravent la gestion urbaine. Ce type d’aléa naturel est de grande actualité, aussi bien sur le plan scientifique que sur le plan médiatique, à cause de l’augmentation de la vulnérabilité, due aux circonstances des changements globaux (réchauffements climatiques) et à la forte urbanisation, souvent irrationnelle. L’objectif de cet article est la mise en place d’une approche objective visant l’évaluation de la susceptibilité aux glissements de terrain dans la ville d’Al Hoceima et sa périphérie. La théorie de l’évidence, qui est une méthode probabiliste bivariée, est fondée sur les règles de Bayes qui consistent à calculer la probabilité d’occurrence spatiale de glissements de terrain, en se basant sur la notion de probabilité à priori et de probabilité à posteriori, tout en considérant les glissements de terrain comme variable à modéliser et chaque facteur causatif comme variable prédictive. Le but de ce travail est de procéder à un zonage d’aléa glissement de terrain tout en assurant une bonne prédiction de ce phénomène avec une bonne résolution spatiale. Les résultats de la courbe de ROC (Receiver Operating Characteristic) montre que la confrontation de la carte de susceptibilité, des glissements de terrain à la carte d’inventaire, permet une capacité de prédiction considérable (AUC=0,889). Ceci pousse au constat selon lequel, plus de 2/3 des glissements de terrain inventoriés s’inscrivent dans des classes de susceptibilité élevée et très élevée. Ce produit cartographique peut constituer un puissant outil d’aide permettant la formulation des suggestions, dans le but d’optimiser l’évaluation du risque de glissements de terrain dans les zones exposées à ce phénomène. Mots clés : SIG, Théorie de l’évidence, Susceptibilité aux glissements de terrain, Al Hoceima (Maroc)

scholarly journals RainForest: A random forest algorithm for quantitative precipitation estimation over Swizerland

2020 ◽  
Author(s):  
Daniel Wolfensberger ◽  
Marco Gabella ◽  
Marco Boscacci ◽  
Urs Germann ◽  
Alexis Berne
Keyword(s):  
Random Forest ◽  
Classical Problem ◽  
Ground Level ◽  
A Posteriori ◽  
Large Database ◽  
Quantitative Precipitation Estimation ◽  
Empirical Relations ◽  
Precipitation Estimation ◽  
Temporal Disaggregation ◽  
Improved Performance

Abstract. Quantitative precipitation estimation (QPE) is a difficult task, particularly in complex topography, and requires the adjustment of empirical relations between radar observables and precipitation quantities, as well as methods to transform observations aloft to estimations at the ground level. In this work, we tackle this classical problem with a new twist, by training a random forest (RF) regression to learn a QPE model directly from a large database comprising four years of combined gauge and polarimetric radar observations. This algorithm is carefully fine-tuned by optimizing its hyper-parameters and then compared with MeteoSwiss' current operational non-polarimetric QPE method. The evaluation shows that the RF algorithm is able to significantly reduce the error and the bias of the predicted precipitation intensities, especially for large and solid/mixed precipitation. In weak precipitation, however, and despite a-posteriori bias correction, the RF method has a tendency to overestimate. The trained RF is then adapted to run in a quasi-operational setup providing 5 minute QPE estimates on a Cartesian grid, using a simple temporal disaggregation scheme. A series of six case-studies reveal that the RF method creates realistic precipitation fields, with no visible radar artifacts, that appear less smooth then the original non-polarimetric QPE, and offers an improved performance for five out of six events.

scholarly journals RainForest: a random forest algorithm for quantitative precipitation estimation over Switzerland

2021 ◽  
Vol 14 (4) ◽  
pp. 3169-3193
Author(s):  
Daniel Wolfensberger ◽  
Marco Gabella ◽  
Marco Boscacci ◽  
Urs Germann ◽  
Alexis Berne
Keyword(s):  
Random Forest ◽  
Classical Problem ◽  
Ground Level ◽  
A Posteriori ◽  
Large Database ◽  
Quantitative Precipitation Estimation ◽  
Empirical Relations ◽  
Precipitation Estimation ◽  
Temporal Disaggregation ◽  
Improved Performance

Abstract. Quantitative precipitation estimation (QPE) is a difficult task, particularly in complex topography, and requires the adjustment of empirical relations between radar observables and precipitation quantities, as well as methods to transform observations aloft to estimations at the ground level. In this work, we tackle this classical problem with a new twist, by training a random forest (RF) regression to learn a QPE model directly from a large database comprising 4 years of combined gauge and polarimetric radar observations. This algorithm is carefully fine-tuned by optimizing its hyperparameters and then compared with MeteoSwiss' current operational non-polarimetric QPE method. The evaluation shows that the RF algorithm is able to significantly reduce the error and the bias of the predicted precipitation intensities, especially for large and solid or mixed precipitation. In weak precipitation, however, and despite a posteriori bias correction, the RF method has a tendency to overestimate. The trained RF is then adapted to run in a quasi-operational setup providing 5 min QPE estimates on a Cartesian grid, using a simple temporal disaggregation scheme. A series of six case studies reveal that the RF method creates realistic precipitation fields, with no visible radar artifacts, that appear less smooth than the original non-polarimetric QPE and offers an improved performance for five out of six events.

scholarly journals Using Different Algorithms and Multi-Seasonal, Textural and Ancillary Information to Increase Classification Accuracy during the Period 2000–2015 in a Mediterranean Semiarid Area

2017 ◽  
Author(s):  
Francisco Gomariz-Castillo ◽  
Francisco Alonso-Sarría ◽  
Fulgencio Cánovas-García
Keyword(s):  
Maximum Likelihood ◽  
Random Forest ◽  
Classification Accuracy ◽  
Maximum A Posteriori ◽  
Support Vector ◽  
Textural Features ◽  
A Posteriori ◽  
Semiarid Area ◽  
Vector Machines ◽  
Terrain Features

The aim of this study is to evaluate three different strategies to improve classification accuracy in a highly fragmented semiarid area. i) Using different classification algorithms: Maximum Likelihood, Random Forest, Support Vector Machines and Sequential Maximum a Posteriori, with parameter optimisation in the second and third cases; ii) using different feature sets: spectral features, spectral and textural features, and spectral, textural and terrain features; and iii) using different image-sets: winter, spring, summer, autumn, winter+summer, winter+ spring+summer; and a four seasons combination. A 3-way ANOVA is used to discern which of these approaches and their interactions significantly increases accuracy. Tukey-Kramer contrast using a heteroscedasticity-consistent estimation of the kappa covariances matrix was used to check for significant differences in accuracy. The experiment was carried out with Landsat TM, ETM, and OLI images corresponding to the period 2000-2015. A combination of four images was the best way to improve accuracy. Maximum Likelihood, Random Forest and Support Vector Machines do not significantly increase accuracy when textural information is added, but do so when terrain features are taken into account. On the other hand, Sequential Maximum a Posteriori increases accuracy when textural features are used, but reduces accuracy substantially when terrain features are included. Random Forest using the three feature subsets and Sequential Maximum a Posteriori with spectral and textural features had the largest kappa values, around 0.9.

Practical Correction of Three Fold Astigmatism in the Philips CM TEM

1996 ◽  
Vol 54 ◽  
pp. 418-419
Author(s):  
Arno J. Bleeker ◽  
Mark H.F. Overwijk ◽  
Max T. Otten
Keyword(s):  
Optical Properties ◽  
Phase Contrast ◽  
The Other ◽  
Objective Lens ◽  
Symmetric Part ◽  
A Posteriori ◽  
Contrast Transfer Function ◽  
High Brightness ◽  
Rotationally Symmetric ◽  
Brightness Field

With the improvement of the optical properties of the modern TEM objective lenses the point resolution is pushed beyond 0.2 nm. The objective lens of the CM300 UltraTwin combines a Cs of 0. 65 mm with a Cc of 1.4 mm. At 300 kV this results in a point resolution of 0.17 nm. Together with a high-brightness field-emission gun with an energy spread of 0.8 eV the information limit is pushed down to 0.1 nm. The rotationally symmetric part of the phase contrast transfer function (pctf), whose first zero at Scherzer focus determines the point resolution, is mainly determined by the Cs and defocus. Apart from the rotationally symmetric part there is also the non-rotationally symmetric part of the pctf. Here the main contributors are not only two-fold astigmatism and beam tilt but also three-fold astigmatism. The two-fold astigmatism together with the beam tilt can be corrected in a straight-forward way using the coma-free alignment and the objective stigmator. However, this only works well when the coefficient of three-fold astigmatism is negligible compared to the other aberration coefficients. Unfortunately this is not generally the case with the modern high-resolution objective lenses. Measurements done at a CM300 SuperTwin FEG showed a three fold-astigmatism of 1100 nm which is consistent with measurements done by others. A three-fold astigmatism of 1000 nm already sinificantly influences the image at a spatial frequency corresponding to 0.2 nm which is even above the point resolution of the objective lens. In principle it is possible to correct for the three-fold astigmatism a posteriori when through-focus series are taken or when off-axis holography is employed. This is, however not possible for single images. The only possibility is then to correct for the three-fold astigmatism in the microscope by the addition of a hexapole corrector near the objective lens.

Expected A Posteriori Estimation of Multiple Latent Traits

2005 ◽  
Author(s):  
Damon U. Bryant ◽  
Ashley K. Smith ◽  
Sandra G. Alexander ◽  
Kathlea Vaughn ◽  
Kristophor G. Canali
Keyword(s):  
A Posteriori ◽  
Latent Traits ◽  
Posteriori Estimation

Implementation of data mining as a support of business application strategy

2018 ◽  
Vol 5 (1) ◽  
pp. 47-55
Author(s):  
Florensia Unggul Damayanti
Keyword(s):  
Data Mining ◽  
Random Forest ◽  
Business Strategy ◽  
Input Parameter ◽  
Data Mining Algorithm ◽  
Complex Data ◽  
Business Decision ◽  
Marketing Department ◽  
Business Application ◽  
Complex Data Sets

Data mining help industries create intelligent decision on complex problems. Data mining algorithm can be applied to the data in order to forecasting, identity pattern, make rules and recommendations, analyze the sequence in complex data sets and retrieve fresh insights. Yet, increasing of technology and various techniques among data mining availability data give opportunity to industries to explore and gain valuable information from their data and use the information to support business decision making. This paper implement classification data mining in order to retrieve knowledge in customer databases to support marketing department while planning strategy for predict plan premium. The dataset decompose into conceptual analytic to identify characteristic data that can be used as input parameter of data mining model. Business decision and application is characterized by processing step, processing characteristic and processing outcome (Seng, J.L., Chen T.C. 2010). This paper set up experimental of data mining based on J48 and Random Forest classifiers and put a light on performance evaluation between J48 and random forest in the context of dataset in insurance industries. The experiment result are about classification accuracy and efficiency of J48 and Random Forest , also find out the most attribute that can be used to predict plan premium in context of strategic planning to support business strategy.

Database-Driven Modeling based on Variable Selection using Random Forest and Its Application for Linear Air Fuel Ratio Sensor Output Prediction

2019 ◽  
Vol 139 (8) ◽  
pp. 850-857
Author(s):  
Hiromu Imaji ◽  
Takuya Kinoshita ◽  
Toru Yamamoto ◽  
Keisuke Ito ◽  
Masahiro Yoshida ◽  
...  
Keyword(s):  
Random Forest ◽  
Variable Selection ◽  
Sensor Output ◽  
Fuel Ratio
Sign in / Sign up

Export Citation Format

Share Document