lets talk about imagery · urban growth habitat loss som abi public safety natural disaster damage...
TRANSCRIPT
ContentImages, Video &
Point Clouds:
Formats
Platforms
Services
Modalities
Mosaic
Datasets
Image
Cubes
Tim
e
Platform
Mapping in
Multiple Dimensions
Base MapsSite Models
Terrain
Analysis
Visual Analysis
MonitoringInspection
Intelligence
Impervious
Surfaces
Finding Spatial
PatternsBuilt
Environments Natural
Environments
Asset
Management
Crop Yields
Urban
Growth
Habitat Loss
SOM
ABIPublic
Safety
Natural
Disaster Damage
Assessment
Economic
Development
Surface
Drainage
Volumetrics
New
Construction
Site
Planning
Ortho
Photos
Land Cover
National
Statistics
Topo Maps
Simulation
Viewsheds
Construction
Progress
The ArcGIS Imagery PlatformTransforming remotely sensed content into information
Search &
Discovery
Simulation
Volumetrics
Creating
a Common Language
Imagery
Lidar
3DVector
Tabular
Real-Time
(IoT)
Big Data
Maps,
Scenes,
Layers
Unstructured
ArcGISIntegrates All Types of Data
BIM
Map Production from Imagery
Site Scan
Ortho Mapping
Ortho Maker
Drone2Map
Stereo
Feature
Capture
DEM
Generation
3D Mesh
Orthophotos
(FUTURE CAPABILITIES)
Automated feature extraction
3D building reconstruction
Drone
Piloting
Drone Fleet
Management
Aerial
Triangulation
3D Point
Clouds
Drone
Flight
Planning
From sensor to accurate derived products
Supporting “any kind of sensor”
“Traditional”
Extended
Orthomapping:• Map Accurate Orthophotos
• Digital Surface Models
Elevation Products:• Point Cloud (LAS)
• 3D Mesh Models
Oblique Photos:• Non-Distorted but Map Accurate
• Complete 3D Measurements*
Overlapping
Static Aerial Photos
2D
3D
Inspection
Example: Drone processing local/Cloud
Data: DataCap OÜ & Archaevision OÜ
Aleksandri kirik - Tartu
ArcGIS Online
Many things happening with “new sensors”
“Traditional”
Extended
Back to “classic” Imagery and EO
150 Formats
Open, Partner and User Content
Satellite
Radar
Multi-
spectral Natural
Color
Lidar
Living Atlas
netCDF/HDF/GRIB World
Elevation
TIF, COG,
MRF, CRF
75 Raster Types
+Python Raster Type
(+Sentinel 3&5)
Sentinel
NAIP
Landsat
As: Local Files, Cloud Storage, Web Services
Aerial
Drones
Video
NITF, JP2,…
Categorical
Terrestrial
Multi-
dimensional
Support for all imagery and raster types
Formats
Platforms
Services
Modalities
MODIS
ContentData and information to make decisions
World Imagery Wayback
World Elevation &Terrain
Landsat 7/8 Sentinel 1/2/3/5
MODIS
Living Atlas of the World
Partners
+ All Your Content
Maxar PlanetVexcel
NearmapCyclomedia
…
…
DIAS Support ✓
Azure Support ✓
Amazon Support ✓
…
The “big issue”
Product No processed No in progress No 1hr No 24hr
Sentinel 1 4.714.000 133 176 3344
Sentinel 2 19.221. 000 260 848 14328
Both are very static snapshots from Thursday, 20th of August 2020, 18:00pm CEST on CREODIAS
A “typical problem”
Count Clouds Average TB ∑ TB
14200 all 0.05 710
4190 < 10% 0.05 210
Check on CREODIAS finder using (this is the 10% version):
https://finder.creodias.eu/resto/api/collections/Sentinel2/search.json?maxRecords=10&
cloudCover=%5B0%2C10%5D&sortParam=startDate&sortOrder=descending&status=all&
geometry=POLYGON((21.246568219215646+59.52774183042649%2C21.228027734848634+
57.693440622187666%2C28.329033247413783+57.594215588620926%2C28.347573731780788+
59.490111762885306%2C21.246568219215646+59.52774183042649))&dataset=ESA-DATASET
~ + 200 TB per month
Sentinel LivingAtlas Service “SentinelView”
Number of scenes (+ 1600 a day)
• Sentinel-2 1c: > 4.000 000
Data Volume (just creating proxies to org-data, holding “best” as mrf)
• LocalCache of Best scenes : ~ 22 TB
• Recycling of cache : every 5 days
• Hosting: Amazon (Frankfurt)
https://sentinel.arcgis.com/arcgis/rest/services/Sentinel2/I
mageServer/query?where=1%3D1&returnCountOnly=true
&f=html
Have a guess for Estonia?
Open data: EC Copernicus program
Your access to
Sentinel 2 services
Evolution: From Scenes to Temporal Stacks
…the Volume, Velocity and Variety of imagery are increasing
Evolution: From Pixels to Multidimensional Image Cubes
Creating an information rich multidimensional view of the world
Pixel Aligned
Image Cube
Transposed
Image Cube
Image files
TIF, netCDF, etc
time
longitude
Mosaic Dataset
Image Cube
Longitude
Tim
e
The history of earth observation data, and evolution of information technology and the internet have
created a transition of scene-based and project-based thinking to imagery as a seamless time series.
“ARD”
Hydrological modelling
Terrain Analysis
Spectral Analysis
Pixel Classification
Distance Modelling
Object Detection
Image Classification
Advanced
Raster Functions
Map Algebra
Object Classification
Instance Segmentation
Change Detection Multidimensional
Analysis
Suitability
Analysis
Image Analysis Extracting Information From Imagery
ArcGIS API
for Python
Image
Analyst
Image
Server
Spatial
Analyst
Open data: EC Copernicus program
Image Cube “Tallinn”
FutureIs all about Images you’ll never see
See 2.000.000.000 km2 of Images
Earth has 510.100.000 km²
Africa Landmass is 30.400.000 km²
https://grassland-mundi.egeos-services.it/MowingViewer/
Evolution: From Pixels to Information Services
The Remote Sensing Value Model
for each application of
Remote Sensing…
Spatial/Spectral Resolution
Temporal Resolution
…there is a state of value
equilibrium that must be achieved
Deep learning
with Imagery
Why is deep learning with imagery important?
Traditional human image interpretation doesn’t scale….
More sensors
Large volumes of imagery
Velocity of data
Automation
Accuracy
End-to-end from raw imagery to structured information products
Deep Learning in ArcGIS
Image
Management
Labelling Data
PrepTrain
Model
Inferencing AnalysisField
Mobility, Monitoring
ArcGIS being used for each step of the deep learning workflow
Deep Learning in ArcGIS
Data Types
LiDAR
Point
CloudText
Satellite
Imagery
Aerial
Imagery
Motion
Imagery
Bathymetric
Data
Oriented
Imagery
IntegrationArcPy
TasksImage
Classification
Segmentation Object
Detection
Pixel
Classification
Super
ResolutionFeature
Attachments
• Super Resolution
• FasterRCNN
• YOLOv3
• PointCNN
• FullyConnectedNetwork
• MLModel
• Tf-lite support
• Deep Learning Installer
New
Austrian Open data
DL “Detect Houses”
Where we offer machine learning integration.
ArcGIS API for Python
ArcGIS Notebooks
ArcGIS Pro
ArcGIS Online
ArcGIS Enterprise
ArcGIS Hub - Citizen Data Science
ArcGIS QuickCapture - Edge AI (in R&D)
ArcGIS Insights
ArcGIS Pro for Intelligence
ArcGIS Analytics for IoT - in R&D
ArcGIS Is A Platform For Big Data Spatial Analytics
Faster (10x+)
Power
Outages
(50+ Million)
Density
Imagery
Lidar:
Bare Earth
Hot Spots
Riparian AreasSpace-Time Cube
Lidar:
First Return
. . . Faster and Massively Scalable
Leveraging Distributed Computing and Parallel Processing
Image ServerLarge Imagery Collection
Raster AnalyticsFeature Analytics
GeoAnalytics ServerLarge Observation Collections
Imagery and Remote SensingArcGIS is a comprehensive imagery platform
Image
Management
Map Production
Analysis & AIVisualization &
Exploitation
Content
Site Scan
Drone2Map
Ortho Mapping
DTM Generation
Pixel Editing
Stereo Feature Capture
All Types
Formats
Platforms
Modalities
Local, Cloud, Services
Raster Processing
Spectral Indices
Classification
Segmentation
Deep Learning
Change Detection
Multidimensional
Pro
Web
Excalibur
Motion Imagery
Oriented Imagery
Mosaic Datasets
Image Cubes
Tiled Imagery
Dynamic Imagery
Visualization & ExploitationEnable human interpretation of imagery & derived data
Web App Builders
Configurable templates
Story Maps
Field Operations apps
ArcGIS Excalibur
ArcGIS Earth
ArcGIS Pro
Oriented Imagery
Motion Imagery
Oblique Imagery
Stereo Imagery
Multidimensional Data
Exploit all kinds of imagery and rasters
Straightforward user experiences for web and desktop
Use imagery to share your story
Make informed decisionsCustom web apps
Guenter Doerffel | Imagery & Raster AnalyticsEsri | Groot Handelsgebouw - Stationsplein 45 | 3013 AK Rotterdam |The Netherlands
T: +31 (0)10 217 77 88 | M: +31 (0)6 83 83 56 15 | [email protected] | esri.com