Optimizing Forensic Damage Investigations for Post ... · Optimizing Forensic Damage Investigations for Post-disaster Operations March 6, 2019 Richard L. Wood J. ArnWomble ... Available

Optimizing Forensic Damage Investigations for Post-disaster

Operations

March 6, 2019

Richard L. WoodJ. Arn Womble

Christine E. Wittich

Motivation for reconnaissance efforts

Introduction to windstorms

Available equipment

Digital resources and planning

Field implementation

Recommended best practices

Outline



Available equipment




Outline

Initial assessments – prioritize efforts / planning

Search and rescue

Emergency management

Preserve damage evidence (before alteration)

Claims information and forensic analysis

Learning from damage (building product performance, wind speeds, wind fields, wind patterns, training)

Pathway to automated damage assessments

Why do we need rapid reconnaissance?

Motivation | Introduction | Equipment | Resources | Implementation | Best Practices 4



Available equipment




Outline

Reconnaissance strategies can depend on

Location (rural vs. urban)

Building stock (residential vs commercial vs industrial)

Size of population affected

Access to damage area

Examples – 2 wind storm scenarios

Introduction to Wind Storms


Tornado – Pampa, TX (2015)

N

Halliburton Oilfield Services

EF-3 damage to steel structures

Isolated rural site

Private industrial facility

Private aerial photographer (3 days)

UAS and lidar (3 weeks)




Lidar collection from fence line due to site access restrictions



UAS imaging – aid to explain and detail a puzzling damage pattern

Reveal criticalbuilding components not visible from thefence line

2 inch aerial imagerypresented here



Garage doors are problematic

Which parts of a warehouse are safest?



Garage doors are problematic

Which parts of a warehouse are safest?



Which parts of a warehouse are safest?(… one more chance)



Which parts of a warehouse are safest?(… one more chance)



Multiple populated areas

Significant public attention

NOAA imagery – Entire TX coast (>360 miles)

Hurricane Harvey (2017)


Hurricane Harvey (2017) “Known” (measured)

wind speedsRockport

FCMP


Hurricane Harvey (2017) Targets identified by visual inspection of NOAA images




Available equipment




Outline

Also known as laser scanning

Active remote sensing technique

Mechanism

Emits pulse or wave of laser light

Laser reflections are detected

Depth to surface calculated

Images overlaid for RGB color

Lidar Mechanism

Available Equipment: Lidar


Yields 3D point cloud

Vertices in 3D space representing reflected surfaces

RGB color and intensity per vertex

Multiple modalities

Terrestrial (ground-based) –presentation focus

Mobile (vehicle-based)

Aerial (plane-based)


Lidar Mechanism


Line-of-sight technology

Potential for significant areas of occlusion in scene

Results in “holes” in point cloud


Occlusion in Point CloudLidar Mechanism


Available Equipment: Lidar Scanning Strategies:

Verify scanner’s field of view

Multiple scans to reduce occlusion

Point clouds must be registered to common reference frame

Increases error due to registration

Ensure scan resolution matches post-disaster needs (mm-level vs. cm-level)


Available Equipment: Lidar Representative Platform:

FARO Focus 3D x130

Range: 130 m

Acquisition Rate: 976,000 pt/s

User-Defined Resolution:

Maximum: 0.15 mm at 25 m

Resolution increases at closer distances FARO Focus 3D


Available Equipment: UAS Unmanned aerial systems for

imaging and point clouds

Line-of-sight technique

Point cloud generated in post-processing

Sensitive to environmental factors and regulations

Wind & precipitation

Pilot licensure

DJI Mavic Pro

UAS in the Field


Available Equipment: UAS Images from UAS can yield similar point clouds

through Structure-from-Motion (SfM)

Detects features in images

Camera location and orientation triangulated

Feature points placed in 3D space

Unscaled - need reference geometry

Structure-from-Motion(Image from Mathworks)


Representative Point Cloud



Available equipment




Outline

Digital Reconnaissance Digital or “virtual” reconnaissance

Use of online media to aid in damage identification

Particularly useful for reconnaissance planning

Locations of heavy damage

Types of structures

Equipment decisions


Digital Reconnaissance Digital sources of information

NWS Damage Survey Viewer

Google News Alerts

Local gov. social media

Multi-hazard applicability

Windstorms/tornadoes

Earthquakes

Floods Digital Sources


Digital Reconnaissance: 2018


Digital Reconnaissance: Results Wide range of damage identified:

Rural (All) Irrigation (Pivot)

Barn or Agricultural Structure Bin or Silo

Crop Equipment

Standard Residential Home Manufactured/ and or Mobile


Digital Reconnaissance: Case Study Digital Recon. Results:

Loc: Nickerson, NE

3 center pivots damaged

1 barn with roof damage

Field Recon. Results: 18 damaged pivots

UAS to identify and map Poor visibility from street

1 barn with roof damage


Digital Reconnaissance: Case Study

Data visualization

Topographical effects

Damage identification

Debris tracking

Global damage assessment

Local damage assessment

Depends on Ground Sampling Distance

2018 Pella, IA EF-3 Tornado

Point Cloud Applications


Tier Approx. GSD1. City ~ 2 cm +2. Neighborhood 1 - 2 cm3. Structure 0.5 - 1.5 cm4. Component 0.5 - 10 mm



Available equipment




Outline

Veterans of Foreign Wars (VFW) building

Aransas Pass storage unit facility

Both lidar and UAS data collection were conducted

Ground control points (GCP) by real-time kinematic (RTK) GPS

Case Study 1: Hurricane Harvey

Map of Rockport, TX


Equipment: DJI Inspire 2, Zenmuse X5 camera and mounted 15 mm lens

VFW: 3 flights 85% overlap at an above-ground-level (AGL) altitude of 25 meters

Aransas Pass storage: 2 flights 85% overlap and an AGL altitude of 28 meters

UAS Platform

VFW aerial image locations

UAS platform


2 lidar scanners:

Faro Focus 3D S350

Faro Focus 3D X330

Six lidar scans at Rockport VFW

Sixteen scans at Aransas Pass storage

Lidar Platform

VFS GCP shown in green


UAS GSD: 0.55-0.61 cm

VFW SfM cloud: 143 million points

Aransas Pass storage SfMcloud: 460 million points

5 distributed GCPs

UAS Processed Data

Storage facility

VFW


UAS Processed Data: Map View


UAS Processed Data: Flight View


UAS Processed Data: GCP View


UAS Processed Data: Orthoimage

UAS Processed Data: Orthoimage

UAS Processed Data: Point Cloud


UAS Processed Data: Point Cloud 2


Use of GCP and natural targets for registration

Mean alignment errors:

0.17 cm VFW

0.47 cm SF

Consistently dense

Lidar Processed Data

Storage facility

VFW


Lack of GCPs show extraneous points

Flatness varies, GCP results in consistency

Data Comparison: UAS Visual

Storage facility

VFWno GCPs

with GCPs

VFW


Model constructed with and w/o GCPs

Errors introduced up to several meters

More pronounced for corridor-like sites (SF here)

Data Comparison: UAS Discrete Errors

with GCPs

Err

or in

Met

ers

Err

or in

Met

ers

no GCPs


Data Comparison: UAS Distributed Errors Cloud-to-cloud comparison with lidar

Lidar reg. error 0.47 cm (0.19 inch)

no GCPs

Dis

tanc

e in

met

ers


Data Comparison: UAS Distributed Errors Cloud-to-cloud comparison with lidar

Lidar reg. error 0.47 cm (0.19 inch)

with GCPs

Dis

tanc

e in

met

ers


Measurement of structural components

Longest wall

GCPs reduce error by an order of magnitude

Errors vary throughout the cloud

Data Comparison: Take-off Quantity

Storage Facility

UAS Length (m) % Diff.No GCP 189.38 1.773%

With GCP 186.28 0.107%


Measurement of structural components

Longest wall

GCPs reduce error by an order of magnitude

Errors vary throughout the cloud

Data Comparison: Take-off QuantityVFW

UAS Length (m) % Diff.No GCP 31.12 0.923%

With GCP 31.44 0.096%




Available equipment




Outline

Detailed equipment options:

Ground-based lidar

Unmanned aerial systems

Equipment deciding factors:

Accessibility and permission

Efficiency (& economical)

Products

Density and accuracy

Best Practice Recommendations


2018 Pella, IA EF-3 Tornado

Factor Lidar UASAccessibility High MediumPermission Yes Depends*

Efficiency Low High

Products Point Clouds

ImagesVideos

OrthomosaicPoint Clouds

Accuracy sub-mm to cm mm to cmDensity mm to cm level cm level

Equipment Deciding Factors


Product Approx. Size UsageImage 5 MB/image Visual damage identificationVideo 100 MB/minute Visual damage identification

Orthomosiac 300-800 MB/site

Visual damage identification2D measurementsElevation contours

Point Cloud 3 – 30 GB/site

Visual damage identification2D and 3D measurements

Elevation contoursMember sizes

Localized deformations

Deliverables Evaluation


Choose appropriate minimum equipment

Verify and seek required permissions

Federal Aviation Administration (FAA)

LEO roadblock

Site and property access 2018 Bondurant, IA Tornado

Recommended Field Guidance


Away from airport (5 miles) and in permissible airspace

Under 400 feet in good weather

In unassisted line-of-sight

Give way to manned aircraft

No flights over non-participants and moving vehicles

Aircraft registration required

Part 107 Basics

General Guidance on UAS Part 107


FAA Safety Regulations

General Guidance on UAS Rules






Pre-mission planning

AirMap

Kittyhawk

Flight planning

Litchi

Pix4Dcapture

DJI Go 4

Available Resources


Available Resources


Pre-mission planning

AirMap

Kittyhawk

Flight planning

Litchi

Pix4Dcapture

DJI Go 4

Lidar and UAS platforms can capture detailed and accurate data

UAS deployments are typically efficient and require less manpower

UAS should be coupled with ground control for accurate and reliable data from structure-from-motion

Forensic studies can be applied with identified damage failure mechanisms and member sizes

Summary


Support from various students and collaborators including: Yijun Liao, M. Ebrahim Mohammadi,

Lianne Brito, Andrew Loken Peter Hughes, Rusty Johnston Halliburton Corporation

Funding resources: National Science Foundation (NSF) Awards:

CMMI-1760010 CMMI-1623553 (RAPID Response – Pampa, TX Tornado) CMMI-1626480 (MRI – Laser Scanner / WTAMU) CMMI-1760010 (RAPID Response – Hurricane Harvey) CMMI-1751018 (CAREER/ Womble)

NSF/Geotechnical Extreme Events Reconnaissance: CMMI-1266418

University of Nebraska System Science Initiative University of Nebraska Foundation

Acknowledgements


Thank you!

Richard L. [email protected]

J. Arn [email protected]

Christine E. [email protected]

Optimizing Forensic Damage Investigations for Post-disaster Operations

Optimizing Forensic Damage Investigations for Post ... · Optimizing Forensic Damage Investigations for Post-disaster Operations March 6, 2019 Richard L. Wood J. ArnWomble ... Available

Documents