Paulo Gonçalves1 Hugo Carrão2 André Pinheiro2 Mário Caetano2

Slides:



Advertisements
Similar presentations
(SubLoc) Support vector machine approach for protein subcelluar localization prediction (SubLoc) Kim Hye Jin Intelligent Multimedia Lab
Advertisements

ECG Signal processing (2)
Image classification Given the bag-of-features representations of images from different classes, how do we learn a model for distinguishing them?
GLC 2000 ‘Final Results’ Workshop (JRC-Ispra, 24 ~ 26 March, 2003) GLC 2000 ‘Final Results’ Workshop (JRC-Ispra, 24 ~ 26 March, 2003) LAND COVER MAP OF.
LAND COVER MAPPING OVER France USING S1-S10 VEGETATION DATASET J-L CHAMPEAUX, S. GARRIGUES METEO-FRANCE GLC 2000 – “FIRST RESULTS” WORKSHOP JRC – Ispra,
A gene expression analysis system for medical diagnosis D. Maroulis, D. Iakovidis, S. Karkanis, I. Flaounas D. Maroulis, D. Iakovidis, S. Karkanis, I.
VEGETATION MAPPING FOR LANDFIRE National Implementation.
Processing methodology for full exploitation of daily VEGETATION data C. Vancutsem, P. Defourny and P. Bogaert Environmetry and Geomatics (ENGE) Department.
Salvatore giorgi Ece 8110 machine learning 5/12/2014
Leila Talebi, Anika Kuczynski, Andrew Graettinger, and Robert Pitt
Desheng Liu, Maggi Kelly and Peng Gong Dept. of Environmental Science, Policy & Management University of California, Berkeley May 18, 2005 Classifying.
Statistical Learning Theory: Classification Using Support Vector Machines John DiMona Some slides based on Prof Andrew Moore at CMU:
CHANGE DETECTION METHODS IN THE BOUNDARY WATERS CANOE AREA Thomas Juntunen.
Module 2.1 Monitoring activity data for forests using remote sensing REDD+ training materials by GOFC-GOLD, Wageningen University, World Bank FCPF 1 Module.
Landscape-scale forest carbon measurements for reference sites: The role of Remote Sensing Nicholas Skowronski USDA Forest Service Climate, Fire and Carbon.
Ramesh Gautam, Jean Woods, Simon Eching, Mohammad Mostafavi, Scott Hayes, tom Hawkins, Jeff milliken Division of Statewide Integrated Water Management.
Co-authors: Maryam Altaf & Intikhab Ulfat
earthobs.nr.no Land cover classification of cloud- and snow-contaminated multi-temporal high-resolution satellite images Arnt-Børre Salberg and.
Global land cover mapping from MODIS: algorithms and early results M.A. Friedl a,*, D.K. McIver a, J.C.F. Hodges a, X.Y. Zhang a, D. Muchoney b, A.H. Strahler.
Chenghai Yang 1 John Goolsby 1 James Everitt 1 Qian Du 2 1 USDA-ARS, Weslaco, Texas 2 Mississippi State University Applying Spectral Unmixing and Support.
Machine Learning1 Machine Learning: Summary Greg Grudic CSCI-4830.
Classification & Vegetation Indices
A METHODOLOGY TO SELECT PHENOLOGICALLY SUITABLE LANDSAT SCENES FOR FOREST CHANGE DETECTION IGARSS 2011, Jul, 27, 2011 Do-Hyung Kim, Raghuram Narashiman,
Fuzzy Entropy based feature selection for classification of hyperspectral data Mahesh Pal Department of Civil Engineering National Institute of Technology.
1 SUPPORT VECTOR MACHINES İsmail GÜNEŞ. 2 What is SVM? A new generation learning system. A new generation learning system. Based on recent advances in.
Image Classification 영상분류
Generating fine resolution leaf area index maps for boreal forests of Finland Janne Heiskanen, Miina Rautiainen, Lauri Korhonen,
The study area is the Sub-Saharian Africa. According to the IGBP vegetation map the major vegetation types present in the area include savanna and woody.
Spectral classification of WorldView-2 multi-angle sequence Atlanta city-model derived from a WorldView-2 multi-sequence acquisition N. Longbotham, C.
North American Croplands Teki Sankey and Richard Massey Northern Arizona University Flagstaff, AZ.
IMPROVING ACTIVE LEARNING METHODS USING SPATIAL INFORMATION IGARSS 2011 Edoardo Pasolli Univ. of Trento, Italy Farid Melgani Univ.
Application of spatial autocorrelation analysis in determining optimal classification method and detecting land cover change from remotely sensed data.
H51A-01 Evaluation of Global and National LAI Estimates over Canada METHODOLOGY LAI INTERCOMPARISONS LEAF AREA INDEX JUNE 1997 LEAF AREA INDEX 1993 Baseline.
GLC 2000 Workshop March 2003 Land cover map of southern hemisphere Africa using SPOT-4 VEGETATION data Ana Cabral 1, Maria J.P. de Vasconcelos 1,2,
U.S. Department of the Interior U.S. Geological Survey Chandra Giri Ying Zhong 7/15/2015 Cropland Extent Mapping in South America.
Computer Vision Lecture 7 Classifiers. Computer Vision, Lecture 6 Oleh Tretiak © 2005Slide 1 This Lecture Bayesian decision theory (22.1, 22.2) –General.
Landsat Satellite Data. 1 LSOS (1-ha) 9 Intensive Study Areas (1km x 1km) 3 Meso-cell Study Areas (25km x 25km) 1 Small Regional Study Area (1.5 o x 2.5.
Integrating LiDAR Intensity and Elevation Data for Terrain Characterization in a Forested Area Cheng Wang and Nancy F. Glenn IEEE GEOSCIENCE AND REMOTE.
US Croplands Richard Massey Dr Teki Sankey. Objectives 1.Classify annual cropland extent, Rainfed-Irrigated, and crop types for the US at 250m resolution.
1 Bilinear Classifiers for Visual Recognition Computational Vision Lab. University of California Irvine To be presented in NIPS 2009 Hamed Pirsiavash Deva.
High resolution product by SVM. L’Aquila experience and prospects for the validation site R. Anniballe DIET- Sapienza University of Rome.
Combining Models Foundations of Algorithms and Machine Learning (CS60020), IIT KGP, 2017: Indrajit Bhattacharya.
Gofamodimo Mashame*,a, Felicia Akinyemia
Temporal Classification and Change Detection
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
ECE 471/571 - Lecture 19 Review 02/24/17.
LAND COVER CLASSIFICATION WITH THE IMPACT TOOL
Mammogram Analysis – Tumor classification
Built-up Extraction from RISAT Data Using Segmentation Approach
HIERARCHICAL CLASSIFICATION OF DIFFERENT CROPS USING
Basic machine learning background with Python scikit-learn
Incorporating Ancillary Data for Classification
Feature Extraction “The identification of geographic features and their outlines in remote-sensing imagery through post-processing technology that enhances.
الدكتور: أحمد رأفت غضية صفاء عبد الجليل كامل حمادة
Supervised Classification
Hyperparameters, bias-variance tradeoff, validation
NASA alert as Russian and US satellites crash in space
Unsupervised Classification
Planning a Remote Sensing Project
ECE 471/571 – Review 1.
Igor Appel Alexander Kokhanovsky
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Other Classification Models: Support Vector Machine (SVM)
WP05: Turkey Land Use Change Model
Machine Learning – a Probabilistic Perspective
Calculating land use change in west linn from
Support Vector Machines 2
Hairong Qi, Gonzalez Family Professor
ECE – Pattern Recognition Midterm Review
Presentation transcript:

LAND COVER CLASSIFICATION WITH SUPPORT VECTOR MACHINE APPLIED TO MODIS IMAGERY Paulo Gonçalves1 Hugo Carrão2 André Pinheiro2 Mário Caetano2 1 INRIA Rhône-Alpes, France 2 Instituto Geográfico Português, Portugal 15/11/2018 25th EARSeL Symposium, Porto 2005

Motivations and objectives Automatic land cover classification, from moderate resolution reflectance imagery MODIS data specificities: - Multi-spectral reflectances: 7 bands (459nm 2155nm) Multi-temporal aquisition: weekly composite of daily images over one year - Spatial resolution: 500 meters - Radiometric corrections and geolocated calibration - Vegetation indices: NDVI and EVI Methodology: Support Vector Machine (SVM) 15/11/2018 25th EARSeL Symposium, Porto 2005

Classes nomenclature and sampling 9 land cover classes based on CORINE Land Cover 2000 nomenclature: - Water areas - Urban areas - Bare soils - Natural grasslands - Shrub lands - Needleleaf forests - Broadleaf forests - Non irrigated lands - Irrigated lands 15/11/2018 25th EARSeL Symposium, Porto 2005

Classes nomenclature and sampling 9 land cover classes based on CORINE Land Cover 2000 nomenclature: - Water areas (15) - Urban areas (30) - Bare soils (15) - Natural grasslands (30) - Shrub lands (30) - Needleleaf forests (30) - Broadleaf forests (30) - Non irrigated lands (31) - Irrigated lands (30) Sample collection based on CORINE Land Cover 2000 cartography and Landsat ETM+ images 15/11/2018 25th EARSeL Symposium, Porto 2005

25th EARSeL Symposium, Porto 2005 MODIS data processing toolbox (Matlab®) 15/11/2018 25th EARSeL Symposium, Porto 2005

Time sliding windowed median filter Window width 25th EARSeL Symposium, Porto 2005

Cloud corrected profiles Band 2 NDVI EVI Band 3 Band 7 Band 4 Band 5 Band 6 15/11/2018 25th EARSeL Symposium, Porto 2005

Support Vector Machine Supervised machine learning system (Vapnik et al.,92) 1) Non-linear mapping of data into higher dimensional space 2) Linear separation in feature space Non-linear classification in input space Exhaustive benchmarks with standard classifiers exist (Nearest Neighbors, Neural Networks, Maximum Likelihood, Classification Trees...) What kind of beast is this??? 15/11/2018 25th EARSeL Symposium, Porto 2005

25th EARSeL Symposium, Porto 2005 Support Vector Machine 15/11/2018 25th EARSeL Symposium, Porto 2005

25th EARSeL Symposium, Porto 2005 Global classification performances Experimental protocol Principal components analysis with full rank representation (9 dim) Cross validation to estimate generalized classification error (5 folds) Single date measurements K-NN SVM Generalized classification errors as functions of time 15/11/2018 25th EARSeL Symposium, Porto 2005

25th EARSeL Symposium, Porto 2005 Class identification accuracy User Producer 15/11/2018 25th EARSeL Symposium, Porto 2005

25th EARSeL Symposium, Porto 2005 Partial conclusions and future work Land Cover Classification Optimal date for MODIS-based Land Cover classification Need to revise Classes Definition and Nomenclature Need to collect new samples sets for Training and for Testing Classification Methodology Promising application of SVM learning systems Fit the times series with (parsimonious) parametric models (for data fusion and dimensionality reduction) Use model parameters as inputs of SVM classifiers 15/11/2018 25th EARSeL Symposium, Porto 2005

Producer's Accuracy (%) Class identification accuracy Classification Reference Data   W NG BLF BS SL NLF IL NIL U User's Accuracy (%) 13 1 92.9 18 2 5 3 52.9 25 86.2 10 76.9 24 61.5 92.6 22 73.3 26 Producer's Accuracy (%) 86.7 60.0 83.3 66.7 80.0 71.0 15/11/2018 25th EARSeL Symposium, Porto 2005