ESTIMATION OF EARTHQUAKE DAMAGE FROM AERIAL IMAGES BY PROBABILISTIC METHOD Shota Izaka, Hitoshi Saji (Shizuoka University)

Slides:



Advertisements
Similar presentations
Applications of one-class classification
Advertisements

Pseudo-Relevance Feedback For Multimedia Retrieval By Rong Yan, Alexander G. and Rong Jin Mwangi S. Kariuki
Context-based object-class recognition and retrieval by generalized correlograms by J. Amores, N. Sebe and P. Radeva Discussion led by Qi An Duke University.
By: Mani Baghaei Fard.  During recent years number of moving vehicles in roads and highways has been considerably increased.
Road-Sign Detection and Recognition Based on Support Vector Machines Saturnino, Sergio et al. Yunjia Man ECG 782 Dr. Brendan.
Human Identity Recognition in Aerial Images Omar Oreifej Ramin Mehran Mubarak Shah CVPR 2010, June Computer Vision Lab of UCF.
Carolina Galleguillos, Brian McFee, Serge Belongie, Gert Lanckriet Computer Science and Engineering Department Electrical and Computer Engineering Department.
Naresh N Spatial Modelling Group RMS India Pvt. Ltd., Noida February 8, 2012 Damage loss estimation of the 2011 Japan tsunami: A case study Co-authors.
Maintainable 3D Models of Cities Gerhard NAVRATIL Rizwan BULBUL Andrew U. Frank Vienna University of Technology Institute of Geoinformation and Cartography.
Leila Talebi, Anika Kuczynski, Andrew Graettinger, and Robert Pitt
Ghunhui Gu, Joseph J. Lim, Pablo Arbeláez, Jitendra Malik University of California at Berkeley Berkeley, CA
Sketch Tokens: A Learned Mid-level Representation for Contour and Object Detection CVPR2013 POSTER.
ICIP 2000, Vancouver, Canada IVML, ECE, NTUA Face Detection: Is it only for Face Recognition?  A few years earlier  Face Detection Face Recognition 
ICME 2008 Huiying Liu, Shuqiang Jiang, Qingming Huang, Changsheng Xu.
Chapter 11 Beyond Bag of Words. Question Answering n Providing answers instead of ranked lists of documents n Older QA systems generated answers n Current.
Workshop on Earth Observation for Urban Planning and Management, 20 th November 2006, HK 1 Zhilin Li & Kourosh Khoshelham Dept of Land Surveying & Geo-Informatics.
1 Learning to Detect Objects in Images via a Sparse, Part-Based Representation S. Agarwal, A. Awan and D. Roth IEEE Transactions on Pattern Analysis and.
Efficient Moving Object Segmentation Algorithm Using Background Registration Technique Shao-Yi Chien, Shyh-Yih Ma, and Liang-Gee Chen, Fellow, IEEE Hsin-Hua.
1 On the Statistical Analysis of Dirty Pictures Julian Besag.
Robust Object Segmentation Using Adaptive Thresholding Xiaxi Huang and Nikolaos V. Boulgouris International Conference on Image Processing 2007.
Self-Supervised Segmentation of River Scenes Supreeth Achar *, Bharath Sankaran ‡, Stephen Nuske *, Sebastian Scherer *, Sanjiv Singh * * ‡
Recognition Of Textual Signs Final Project for “Probabilistic Graphics Models” Submitted by: Ezra Hoch, Golan Pundak, Yonatan Amit.
CS 223B Assignment 1 Help Session Dan Maynes-Aminzade.
1 Integration of Background Modeling and Object Tracking Yu-Ting Chen, Chu-Song Chen, Yi-Ping Hung IEEE ICME, 2006.
Student: Hsu-Yung Cheng Advisor: Jenq-Neng Hwang, Professor
Feature Screening Concept: A greedy feature selection method. Rank features and discard those whose ranking criterions are below the threshold. Problem:
Multi-camera Video Surveillance: Detection, Occlusion Handling, Tracking and Event Recognition Oytun Akman.
Redaction: redaction: PANAKOS ANDREAS. An Interactive Tool for Color Segmentation. An Interactive Tool for Color Segmentation. What is color segmentation?
Smart Traveller with Visual Translator for OCR and Face Recognition LYU0203 FYP.
California Car License Plate Recognition System ZhengHui Hu Advisor: Dr. Kang.
Jacinto C. Nascimento, Member, IEEE, and Jorge S. Marques
Introduction to Machine Learning Approach Lecture 5.
© 2013 IBM Corporation Efficient Multi-stage Image Classification for Mobile Sensing in Urban Environments Presented by Shashank Mujumdar IBM Research,
Sensys 2009 Speaker:Lawrence.  Introduction  Overview & Challenges  Algorithm  Travel Time Estimation  Evaluation  Conclusion.
FEATURE EXTRACTION FOR JAVA CHARACTER RECOGNITION Rudy Adipranata, Liliana, Meiliana Indrawijaya, Gregorius Satia Budhi Informatics Department, Petra Christian.
earthobs.nr.no Land cover classification of cloud- and snow-contaminated multi-temporal high-resolution satellite images Arnt-Børre Salberg and.
2 Outline Introduction –Motivation and Goals –Grayscale Chromosome Images –Multi-spectral Chromosome Images Contributions Results Conclusions.
EADS DS / SDC LTIS Page 1 7 th CNES/DLR Workshop on Information Extraction and Scene Understanding for Meter Resolution Image – 29/03/07 - Oberpfaffenhofen.
1. Introduction Motion Segmentation The Affine Motion Model Contour Extraction & Shape Estimation Recursive Shape Estimation & Motion Estimation Occlusion.
ASPRS Annual Conference 2005, Baltimore, March Utilizing Multi-Resolution Image data vs. Pansharpened Image data for Change Detection V. Vijayaraj,
BUILDING EXTRACTION AND POPULATION MAPPING USING HIGH RESOLUTION IMAGES Serkan Ural, Ejaz Hussain, Jie Shan, Associate Professor Presented at the Indiana.
The University of Texas at Austin Vision-Based Pedestrian Detection for Driving Assistance Marco Perez.
1 Multiple Classifier Based on Fuzzy C-Means for a Flower Image Retrieval Keita Fukuda, Tetsuya Takiguchi, Yasuo Ariki Graduate School of Engineering,
Satellite and Aerial Image Analysis. Remote Sensing Earth Observation Photogrammetry From the Cold War to Spaceship Earth Application Areas: anything.
Xu Huaping, Wang Wei, Liu Xianghua Beihang University, China.
Image Segmentation and Edge Detection Digital Image Processing Instructor: Dr. Cheng-Chien LiuCheng-Chien Liu Department of Earth Sciences National Cheng.
Bo QIN, Zongshun MA, Zhenghua FANG, Shengke WANG Computer-Aided Design and Computer Graphics, th IEEE International Conference on, p Presenter.
Prostate Cancer CAD Michael Feldman, MD, PhD Assistant Professor Pathology University Pennsylvania.
Computer-based identification and tracking of Antarctic icebergs in SAR images Department of Geography, University of Sheffield, 2004 Computer-based identification.
Identification and Enumeration of Waterfowl using Neural Network Techniques Michael Cash ECE 539 Final Project 12/19/03.
The Cyber-Physical Bike A Step Toward Safer Green Transportation.
Vehicle Detection in Aerial Surveillance Using Dynamic Bayesian Networks Hsu-Yung Cheng, Member, IEEE, Chih-Chia Weng, and Yi-Ying Chen IEEE TRANSACTIONS.
Cell Segmentation in Microscopy Imagery Using a Bag of Local Bayesian Classifiers Zhaozheng Yin RI/CMU, Fall 2009.
Rescue Robots A social relevant application Arnoud Visser DOAS Kick-off 7 January 2008.
1 A Statistical Matching Method in Wavelet Domain for Handwritten Character Recognition Presented by Te-Wei Chiang July, 2005.
Potential impacts of map error on land cover change detection Nick Cuba Clark University 2/25/12 1.
Date of download: 7/8/2016 Copyright © 2016 SPIE. All rights reserved. A scalable platform for learning and evaluating a real-time vehicle detection system.
Traffic Sign Recognition Using Discriminative Local Features Andrzej Ruta, Yongmin Li, Xiaohui Liu School of Information Systems, Computing and Mathematics.
DIGITAL SIGNAL PROCESSING
Supervised Time Series Pattern Discovery through Local Importance
Gait Analysis for Human Identification (GAHI)
Using aerial images for urban planning
A New Approach to Track Multiple Vehicles With the Combination of Robust Detection and Two Classifiers Weidong Min , Mengdan Fan, Xiaoguang Guo, and Qing.
Lecture 25: Introduction to Recognition
Vehicle Segmentation and Tracking in the Presence of Occlusions
Outline Announcement Texture modeling - continued Some remarks
Aline Martin ECE738 Project – Spring 2005
Image Segmentation.
An introduction to Machine Learning (ML)
Presentation transcript:

ESTIMATION OF EARTHQUAKE DAMAGE FROM AERIAL IMAGES BY PROBABILISTIC METHOD Shota Izaka, Hitoshi Saji (Shizuoka University)

Introduction

Backgrounds After large-scale earthquake –Urban areas are seriously damaged –Many people require rescuing and aid For effective rescue and victim support –Rapid action is needed –A wide range of information is important Aerial images are suitable for disaster observation

Conventional method Matching analysis –Comparing pre-disaster and post-disaster images Difficulty of matching analysis –Difficult to obtain pre-disaster images –Affected by shooting conditions and time Changes of shadows Construction and destruction of buildings

Our goals Rapid analysis of damage –Use only post-disaster aerial images –Not using the training data Assisting various rescue and victim support activities –Providing information available for various purposes Assisting human decisions

Ways of assisting human decisions Remaining undetermined regions –We don’t force to classify all regions –The final decision is left to the people in the field Showing the likelihood of damages –The result available for various purposes Target area estimation of rescue activity Determination of the road passable for emergency vehicles

Method

Overview Aerial Image Segmentation Feature extraction Result for buildings Digital map Region classification Result for roads Road mask creation

Creating road mask from digital map –Roads change little over time Our method is not affected by the time when the map is created Digital mapRoad mask

Segmentation Initial Segmentation –Segment into small basic regions Unification of similar regions –Considering color and textures –Avoiding to unify roads and buildings Before segmentationAfter segmentation

Feature extraction Collapsed buildings –Segmented into small regions –Having short random edges Extracting short edges as a feature of damages Collapsed buildingsSegmented regionsEdges

Feature extraction Undamaged buildings –Maintaining their shapes –Having a large area Extracting building regions as a feature of undamaged Undamaged buildingsSegmented regions Edges

Region classification Using the probabilistic relaxation method –Labeling method using the probability We use the method to classify each region by damage probability

Defining initial probability Considering extracted features –The proportion of short edges –The area of region –Building region or not Large area Building High short edge rate Probability definitions

Probability update Update using similarity –Considering the region similar to damaged region as damaged region Probability update model Low High

Extracting undamaged regions Regions are converged high or low probability Extracting low probability regions as undamaged regions –Considering regions not converged as undetermined regions High probability Result of extraction Low probability Undetermined

Extracting damaged regions from high probability regions High probability Damaged regions extraction model Low probability Undetermined Damaged Undetermined

Redefining initial probability Redefining probability by randomness of edges –Using variance of edge angles Edge model of undamaged buildings Edge model of collapsed buildings

Result of classification ■:Undamaged regions ■:Undetermined regions 1 –Low risk of damage ■:Undetermined regions 2 –High risk of damage ■:Damaged regions Result of classification Undetermined Damaged Undetermined Undamaged

Image division Dividing a result image into buildings and roads –Result of buildings Estimation of building damages –Result of roads Determination of road passable

Experiment

Data Aerial images –Great Hanshin Earthquake –Captured on January 18, 1995 –Provided by PASCO Corp. Digital map –A topographic map of Kobe city –Provided by Kobe City Urban Planning Bureau

Result of classification for buildings Input imageResult image ■:Undamaged regions■:Undetermined regions 1 ■:Undetermined regions 2■:Damaged regions

Result of classification for roads Input imageResult image ■:Undamaged regions■:Undetermined regions 1 ■:Undetermined regions 2■:Damaged regions

Evaluation of accuracy Creating answer images –Using visual judgment Comparing with results Result of classification Undetermined Answer DamagedUndamaged Damaged Undamaged Undetermined

Detection rate Evaluating pixels in same category Result of classification Answer DamagedUndamaged Damaged Undamaged Damaged Undamaged Damaged Undamaged

Detection rate with human decisions Estimating rate after human decisions –Adding undetermined regions Result Damaged Undamaged Answer DamagedUndamaged Damaged Undamaged Damaged Undamaged

False detection rate Evaluating pixels in wrong category –Visual judgment Considered undamaged regions Damaged Undamaged Considered damaged regions Result of classification Damaged Undamaged

Answer for buildings Result imageAnswer image ■:Undamaged regions■:Undetermined regions 1 ■:Undetermined regions 2■:Damaged regions

Answer for roads Answer imageResult image ■:Undamaged regions■:Undetermined regions 1 ■:Undetermined regions 2■:Damaged regions

Result of accuracy evaluation in buildings Undamaged regions –Detection rate:77.2% With human decisions:93.1% –False detection rate:10.1% Damaged regions –Detection rate:74.0% With human decisions:87.0% –False detection rate:17.7%

Result of accuracy evaluation in roads Undamaged regions –Detection rate:85.5% With human decisions:93.4% –False detection rate:19.0% Damaged regions –Detection rate:65.3% With human decisions:79.6% –False detection rate:14.6%

Review of results Obtained high detection rates –Except for damaged regions in roads Features of damage on roads are unclear –Many regions classified into “Undetermined” Requiring human decisions Road imageResult of classification

Review of results Obtained low false detection rates –Roads have more errors than buildings Caused by objects on roads –Cars, roofs, shadows of buildings Roof and carError Shadow and carError

Conclusion Our results can be used for various rescue and victim support activity –Estimation of building damages –Determination of road passable Our future directions –Improving building detection –Detecting objects on roads

End

The Sendai earthquake Most of the damage was caused by the Tsunami Most of the buildings are flooded out –Our method aim to detect collapsed buildings Huge area of damage –Not possible to capture by aerial images Applying to the earthquake is future works