Object-oriented image analysis methods in disaster risk management Dr. Norman Kerle

Transcription

1 Object-oriented image analysis methods in disaster risk management Dr. Norman Kerle

2 Lecture outline (1) ITC (3) Recent research (2) OOA for DRM Natural Hazard Center University of Colorado (4) Outlook & trends 2

3 (1) ITC ITC/ University Twente International Institute for Geoinformation science and Earth Observation independent for 60 years Faculty of University of Twente 3 years ago Various degree courses and certificates in Disaster management Earth sciences Geoinformatics Governance Land administration Natural resources Urban planning Water resources Student numbers Netherlands Indonesia 3

4 ITC/University Twente Houses the United Nations University- ITC Centre for Spatial Analysis and Disaster Risk Management Training, education and curriculum development Knowledge development and research collaboration Advisory services In collaboration with many partners 4

5 ITC Geographer with study background in Hamburg (D), Ohio State (US) and Cambridge (UK) Since early 1990s work in the hazards & disaster field, shortly after focus on remote sensing PhD in volcano remote sensing (lahars) Advanced image analysis, and focus on object-oriented analysis in last 5 years 5

6 (2) OOA for DRM Object-oriented analysis for disaster risk management Disaster risk Different concepts Expected losses (f[hazard, period]) DRM OOA Risk = Hazard * Vulnerability EaR * Amount (R=H*V) EaR (elements at risk): not only physical H: f(type, magnitude) V: physical, social, economic, environmental, etc. Amount: quantifiable? Note: all elements of risk are spatial 6

7 Basics of object-oriented analysis OOA DRM OOA Objects = segments (segmentation-based analysis, OBIA, GEOBIA) 1. step: segmentation: old concept (~1970s) partition an image into homogenous units 2. step: classification of those units 7

8 Basics of object-oriented analysis DRM OOA OOA Segmentation also at multiple scales, and using auxiliary information Super-objects Classification level Sub-objects Pixel 8

9 Basics of object-oriented analysis OOA DRM OOA Main difference over pixel-based methods: objects have extra features (spectral, geometric, contextual) = useful for classification Allows use of feature and process knowledge 9

10 Basics of object-oriented analysis E.g., land use vs. land cover, or process type Google Earth Google Earth Challenge: all depends on hazard type and magnitude (in turn informs EaR and their respective vulnerabilities) 10

11 OOA for DRM aviation hazard Question: do we always understand the hazard, its spatio-temporal characteristics, and its effects? Example: volcanic hazard gases into atmosphere ash/tephra acid rain pyroclastic flow bombs dome collapse possible crater lake outbreak possible glacier melting gases fumaroles All sub-hazards have their own characteristics magma flow lahar/ debris flow Many specific hazards exist flank collapse/ landslide gases cracks volcanotectonics 11

12 (3) Recent research OOA for DRM Our group addresses Use of OOA for different hazards and risk elements Different aspects of risk Methodological work (better segmentation, feature and threshold selection) Domain focus Remote sensing for DRM Hazard Vulnerability EaR Risk Damage (Recovery) Technical focus Landslides/ erosion Social Urban/ infrastructure Pictometry-/UAVbased damage Refugee camps; metrics for recovery OOA 12

13 Hazard: Landslide work Work with several PhD students and postdocs work of Tapas Martha conceptualization of a landslide segmentation based on satellite data and elevation data removal of false positives classification of different landslide types OOA-based landslide mapping Martha et al., 2010 (Geomorphology) Full PhD thesis: 13

14 Hazard: Landslide work Scale factor 20 Scale factor 50 Problem: trial & error work What segmentation parameters? Work on statistical optimization of segmentation Balancing intra-segment homogeneity and inter-segment heterogeneity Plateau objective function (POF) to select appropriate scale factors OOA-based landslide mapping Martha et al., 2010 (Geomorphology) Objective segmentation (POF) Martha et al., 2011 (IEEE TGRS) 14

15 Hazard: Landslide work Ping Lu (Uni Florence): OOA-based landslide change detection Also focused on multi-scale segmentation optimization Pre-event image Post-event image Landslide map OOA-based landslide mapping Martha et al., 2010 (Geomorphology) Objective segmentation (POF) Martha et al., 2011 (IEEE TGRS) Change detection Lu et al., 2011 (IEEE GSRL) 15

16 Hazard: Landslide work Andre Stumpf: classification parameter and threshold selection How to chose from hundreds of object features and the best threshold? Random Forest method (data mining/active learning based on samples) Identifies ideal samples to query, in optimal location Reduces # features Introduces topographicallyguided texture measures 16

17 Hazard: Landslide work Tested on air- and spaceborne data of 4 different sites Accuracies of 73-87% OOA-based landslide mapping Martha et al., 2010 (Geomorphology) Objective segmentation (POF) Martha et al., 2011 (IEEE TGRS) Change detection Lu et al., 2011 (IEEE GSRL) Objective parameter selection Stumpf & Kerle, 2011 (RSE) 17

18 Hazard: Landslide work Tapas Martha: OOA-based landslide detection based only on pan-chromatic data Again use of POF Focus on texture measures Time-series analysis Landslides 18

19 Hazard: Landslide work Objective segmentation (POF) Martha et al., 2011 (IEEE TGRS) Change detection Lu et al., 2011 (IEEE GSRL) Objective parameter selection Stumpf & Kerle, 2011 (RSE) Use of panchromatic data Martha et al, 2012 (ISPRS) 19

20 Hazard: Landslide work with LiDAR data So fall all work focused on optical data Miet Van Den Eeckhaut: detection of forested landslides in single-pule LiDAR data No use of additional optical data Area in Flanders, Belgium; > 200 old deep-seated and shallow slides Almost impossible to detect in optical data (Elevation exaggeration Earth) Rotational slide Complex slide 20

21 Hazard: Landslide work with LiDAR data Procedure: Creation of LiDAR derivatives Multiple segmentation based on POF Detection of main scarp Downslope growing using evidence from side and base scarp, as well as interior Good detection of deep slides (71% of main scarps, >50% of associated landslide body

22 Hazard: Landslide work with LiDAR data Expert Automatic

23 Hazard: Landslide work with LiDAR data Promising results given the challenging terrain Currently plans with Italian colleagues to use radar data (InSAR, permanent scatterer) for landslide mapping and monitoring OOA and LiDAR Van Den Eeckhaut et al., 2012 (Geomorphology)

24 Hazard: Erosion detection Shruthi Rajesh: use of high-resolution satellite data to map gully erosion Similar approaches to what we developed for landslides (directional texture, etc.) Removal of false positives was challenging

25 Hazard: Erosion detection Gully erosion detection Shruthi et al., 2011 (Geomorphology)

26 Hazard: Erosion detection Change detection for gully systems ( ) Gully system change detection Shruthi et al., in press (Catena)

27 Other risk aspects Elements at risk Janak Joshi: Problem - building extraction from optical satellite data Chicken & egg: we d like to have a DEM/DSM, but photogrammetry is imperfect Z X, Y Solution: Create an (imperfect) DEM/DSM Use in OOA (distinguish buildings from similar looking low features) Use the extracted buildings to correct the DEM/DSM 27

28 Other risk aspects Elements at risk Geoeye image Initial DSM OOA-derived buildings Evident errors Assignment of height Corrected DSM 28

29 Other risk aspects Elements at risk Improved DSM, useful for example for flood modeling 29

30 Other risk aspects Social vulnerability Annemarie Ebert: Social vulnerability (SV): people s differential incapacity to deal with hazards, based on the position of the groups and individuals within both the physical and social worlds (Clark et al., 1998) Traditionally assessed using census data (that often don t exist) Solution: use physical proxies We selected 47 variables In stepwise multiple regression against census-based SV index found 8 variables that explained 60% of the variance 30

31 Other risk aspects Social vulnerability SV mapping with OOA Ebert et al., 2009 (Natural Hazards) 31

32 Other risk aspects Deprivation We use similar approaches to map deprivation (e.g. slums) Divyani Kohli: use of spatial metrics to describe urban units extracted with OOA Ontology used to formalize (local/specific) knowledge of slums Currently being used to parameterize OOA-based slum detection Ontology-based slum detection Kohli et al., 2012 (CEUS) OOA-based slum detection Kohli et al., in review (Remote Sensing) 32

33 Other risk aspects Deprivation 33

34 Other risk aspects Damage mapping More recent work Image-based damage mapping per se is very difficult Very little (and not very convincing) OOA work (Pham et al., in review) 34

35 Other risk aspects Damage mapping Our approach (with Markus Gerke): use of oblique image data from Pictometry 5 perspectives, in principle allowing comprehensive damage evaluation Example from Haiti (2010) 2010 Pictometry International Corp 35

36 Other risk aspects Damage mapping Many features were calculated from the data (digital elevation model, texture, etc.) 2010 Pictometry International Corp 36

37 Other risk aspects Damage mapping Images were segmented (OOA) and classified based on training samples Intact roof Broken roof/ rubble Intact facade Bare ground Vegetation Damage mapping with Pictometry data Gerke & Kerle, 2011 (PE&RS) 37

38 (4) Outlook & trends Outlook & trends Status quo: OOA has proven a versatile and useful image analysis concept In DRM it allows effective use of process and feature knowledge Good progress in automating methods High dependence on ecognition (>50% of all papers; high cost) Faster development in the life sciences field than in Earth sciences 38

39 (4) Outlook & trends Outlook & trends Trends & research needs: Many hazard types as yet not addressed Nothing yet for recovery and reconstruction Remaining subjectivity Work on better metrics 39

40 (4) Outlook & trends Outlook & trends 3D data processing 3D data are becoming increasingly available (LiDAR point clouds, geophyscial data) Many inspiring developments from biomedical field. Example: > 500 CT slices of a mouse ecognition result: 40

41 (4) Outlook & trends Outlook & trends 3D data processing Islam Fadel: OOA-processing of geophysical data (seismic/ tomographic data) OOA of 3D geophysical data Fadel et al., in review (Computers & Geosciences) 41

42 (4) Outlook & trends Questions? Our OOA work continues check for updates Same for full references Papers also on Or Thank you 42