Bai

Geo-spatial Data Analysis, Quality Assessment and

Visualization

Yong Ge Hexiang Bai Sanping Li

Institute of Geographic Science &Natural Resources Research, Chinese Academy of Sciences, [email protected] 6/31/2008

What we have faced in geomatics engineering?

in situ sensing

Multi-level Data AcquisitionMulti-platform, multi-sensor, multi-resolution/scales, multi-formats, multi-owners and multi-temporal data

Landuse

DEM

We are drowning in data but starving for knowledge

Geospatial Data

Information

Knowledge

GIS M

odeling

Data --Knowledge

HEC-RAS

Runoff DEM and River River Bank

OpenHydro ArcGIS

Flood Plain Mapping

E

rro

r p

rop

agat

ion

DBMS DBMS DBMS DBMS DBMSData

Prepro

cessing

SDA

Output

Input

Spatial Data Analysis

and

Uncertainty Analysis

Error

Uncertainty

Multi-source Data Acquisition

Spatial Data Analysis

Data Quality Assessment

Output and visualization

Multi-source Data Preprocessing

Strategies

Uncertainty Analysis---Types

Randomness ── Uncertainty of Random phenomenon

Fuzziness ── Uncertainty of Fuzzy concept

Roughness ── Uncertainty of Knowledge in

information system

BG SD SM UA FP

Randomness, Fuzziness and Roughness in SDA

SDA

GIS data sources Evidence maps Modeling

Posterior Prob.

Fuzzy boundary

Random error

Missing data

Randomness

Fuzziness

Roughness

Error factorsLimitations of data acquisition

& processing

Uncertainty Analysis

Input

(Geospatial data)

Output

(Decision +

Uncertainty Analysis)

Population

….

Temperature Data

River Gauge Station

DEM

SDA

Geospatial data

Geospatial data

with error

Environment disturbance

Fuzziness of definition

Knowledge incompleteness

Uncertainty measurement

Accuracy assessment

Error propagation in spatial data

Exploring and weakening uncertaintyError propagation in GIS operation

Error propagation in multi-source data integration

Uncertainty visualization

Integration SDSS with uncertainty

Uncertainty in the integration of GIS and remote sensing

Uncertainty analysis in remote sensing information

Attribute and positional uncertainty analysis

Sensitivity analysis

Uncertainty analysis in spatio-temporal data

Assessment of data quality for original data Accuracy of attribute and position

Raw data

Error propagation models for data preprocessing Assessment of data quality for preprocessed data Accuracy of attribute and position

Preprocessing

Error propagation models for data preprocessing Assessment of data quality for preprocessed data Accuracy of attribute and position

Data analysis

Quality assessment of outcome Accuracy of attribute and position….

Result from spatial data analysis

Multi-level Geo-spatial Data Acquisition

Land surveying: land use and soil type

Land surveying: land use and soil type

Photogrammetry and Remote sensingPhotogrammetry and Remote sensing

Mobile mapping and Mobile surveying

Mobile mapping and Mobile surveying

Measured data: TINs, hydrography, stream stage and precipitation

Measured data: TINs, hydrography, stream stage and precipitation

Geo-spatial Data Preprocessing

Atmospheric Correction, Geometric Correction and Image enhancementAtmospheric Correction, Geometric Correction and Image enhancement

Data format conversion, Projection transformation, Map edit

Data format conversion, Projection transformation, Map edit

Geo-spatial Data Analysis

Classification: Statistics Methods, Neural Methods and Knowledge-

based Methods

Classification: Statistics Methods, Neural Methods and Knowledge-

based Methods

Map operations: AND, OR and MULTIPLEMap operations: AND, OR and MULTIPLE

EDA, Spatial autocorrelation, Fusion, Integration and Information

extraction

EDA, Spatial autocorrelation, Fusion, Integration and Information

extraction

Quality assessment

Geo-spatial data visualization

Geo-spatial data analysis, quality assessment and visualization

Error propagation will

accompany with each

sub-process

Uncertainty analysis and

measurement will accompany

with each sub-process as well

Uncertainty Analysis---Measurement

Existing Methods

There are two fashions to measure attribute

uncertainty of classified remotely sensed imagery:

(1)categorical scale;(2) pixel scale

Three Levels Measurement on Classified Remotely Sensed Imagery

Pixel Class/Object Image

BG SD SM UA FP

Measurements for Classified Pixels

Assess the uncertainty of every classified pixel

Measurement indexes: Shannon Entropy ， Fuzzy Entropy, probability residual ， confusion index ，…

Shannon Entropy：

n

iii ppxH

1

log)(

x is a pixel and n is the number of class types

BG SD SM UA FP

Rough degree of rough set X

Rough entropy of rough set X

The more rough degree/entropy is approaching to 0, the smaller the uncertainty of X is.

||||||||

11))((XXPPXXPP

XXAA 11))((00 XXAA

))||||

1log

||||

||||)()((())((

11

mm

ii ii

iiAAAA

RUUR

XXE ||||log))((00 UXEAA

Measurements for Classes or Objects

BG SD SM UA FP

Accuracy of approximation

Quality of

approximation

Measurements for an Image

n

ii

n

ii

P

XP

XP

1

1

||

||)(

||

||)( 1

U

XPn

ii

P

BG SD SM UA FP

Example for measurement of uncertainty

The study area was selected from a Landsat TM image which was taken over the Chinese Yellow River Delta on August 8, 1999.

BG SD SM UA FP

Classification from MLC

BG SD SM UA FP

1.00

Shannon Entropy

0.00

BG SD SM UA FP

0Degree of Uncertainty

1 (0.00, 0.20]2 (0.20, 0.40]3 (0.40, 0.60]4 (0.60, 0.80]5 (0.80, 1.00]

Uncertainty for Pixels

Rough Degree

0.0000

0.2000

0.4000

0.6000

0.8000

1.0000

1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60

Threshold

Water

Agriculture_1

Agriculture_2

Urban

Bottomland

Bareground

Rough Entropy

0.0000

0.2000

0.4000

0.6000

0.8000

1.0000

1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60

Threshold

Water

Agriculture_1

Agriculture_2

Urban

Bottomland

Bareground

Uncertainty Measurements for Classes or Objects

BG SD SM UA FP

α=1.0

α=0.8

Rough Degree

α=0.6

BG SD SM UA FP

Rough Entropy

α=1.0

α=0.8

α=0.6

BG SD SM UA FP

0.0000

0.2000

0.4000

0.6000

0.8000

1.0000

1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60

Threshold

Quality of Approximation Accuracy of Approximation

BG SD SM UA FP

Uncertainty Measurements for Image

Uncertainty Analysis---Visualization

Uncertainty Visualization

BG SD SM UA FP

NumberVersus

Graph and Map

Visualization--- Static Visualization

3D visualization of uncertainty in water class

BG SD SM UA FP

Visualization--- Dynamic Visualization

BG SD SM UA FP

Class of water: pixels with low uncertainty degree in PCP

Class of water: pixels with high uncertainty degree in PCP

Visualization--- Feature Visualization Sample data clustering in 3D

space

Bottomland

Water

Urban

Agriculture_1

Agriculture_2

Bareground

BG SD SM UA FP

Probability DistributionProbability Distribution

Software development--RASRS – a reliability assessment system for remote sensing information

Thanks!

Explanation of Threshold

NEGATIVENEGATIVE

BoundaryBoundary

Lower

Approximation

Lower

Approximation

Class(x)=i

P(Class(x)=i) > Threshold => Lower

P(Class(x)=i) > 1-Threshold => Upper

P(Class(x)=i) < 1-Threshold => Upper

0<=Threshold<=0.5

Class(x)=i

P(Class(x)=i) > Threshold => Lower

P(Class(x)=i) > 1-Threshold => Upper

P(Class(x)=i) < 1-Threshold => Upper

0<=Threshold<=0.5