state-of-the-art within radiometric correction of large ... · •radiometry: the measurement of...

Post on 28-Jul-2020

6 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

State-of-the-art within

radiometric correction of

large-format aerial

photogrammetric images

Eija Honkavaara, Finnish Geodetic Institute

fagdag Omløp 8.mai 2014

Finnish Geodetic Institute (FGI)

• A research and expert institute that carries out research and development for spatial data infrastructures.

• 4 research departments

• 100 employees

2

Director General

Prof. Jarkko Koskinen

Geodesy and geodynamics

Prof. Markku Poutanen

Geoinformatics and cartography

Prof. Tapani Sarjakoski

Fotogrammetry and remote sensing

Prof. Juha Hyyppä

Navigation and positioning

Prof. Ruizhi Chen

Administration and finances

Ms. Päivi Koponen

Contents

• Introduction

• Theoretical background

• Radiometric sensor calibration

• Radiometric processing

• EuroSDR project 2008-2011

• Examples

• Tree species classification with rigorous radiometric processing

• NDVI time series in Catalonia

• Impact of solar elevation in photogrammetric processes

• Conclusions

3

Introduction • The digital large-format photogrammetric sensors provide excellent

radiometric potential: • Great image quality: Linearity, High dynamic range

• Multispectral data: Photogrammetric remote sensing

• Proper utilization of radiometry provides significant advantages for photogrammetric processes

• Efficient image post-processing

• More flight hours (lower solar altitude)

• Improved visual image quality

• Reflectance calibration

• Improved performance and automation potential of conventional applications

• New quantitative applications, high automation level: classification, monitoring, change detection, …

• Rigorous radiometric processing is a new issue in photogrammetric processing lines and requires new developments

4

Digital photogrammetric cameras DMC

ADS

UltraCam

Microsoft, 2007

Leica Geosystems, 2007

Intergraph, 2007

Intergraph, 2012

DMC2

5

Other relevant airborne imaging

sensors

6

G

R

G

R

B

G

B

G

G

R

G

R

B

G

B

G

Trimble DSS

Small-format customer camers Hyperspectral sensors

© SPECIM Oy Applanix, 2007

Oblique cameras

e.g. Leica

RCD30 Oblique

Penta Pod

DMC orthophoto mosaic

Factors influencing DN

Atmosphere

Illumination

Object

System

7

Theoretical background

8

Definitions • Radiometry: the measurement of

optical radiation in wavelength

range of 0.1-1000 μm

• Radiance

• Measure of the quantity of radiation that

falls from a surface within a given solid

angle in a specified direction

• SI-unit [W m-2 sr-1]

• Photometry: the measurement of

light weighted by the response of

the human eye

• Quantitative remote sensing: Image

DNs are transformed to reflectance

signatures and quantitatively

analyzed

9

Spectral band Wavelength

range

Visible (v) 0.4-0.7 μm

Near infrared (NIR) 0.7-1.1 μm

Short Wave Infrared

(SWIR)

1.1-2.5 μm

Mid Wave Infrared

(MWIR)

3.0-5.0 μm

Thermal or LongWave

Infrared (TIR or LWIR)

8.0-14 μm

At-sensor radiance

Lat_sensor = Ls + Lsky + Lbg+ Lbg_multi+Ladj + Latm.

Ls : direct sun illumination Lsky : Skylight Lbg : Background Lbg_multi: Background multiple Ladj : Adjacency effect Latm : Atmospheric path radiance

/cos)()(),,,,(),,,,( 0

svsrriirriis EL

At-sensor radiance

Direct sun component

At-sensor radiance: radiance

entering to the imaging system

Radiation components

• MODTRAN4 at-sensor

radiance simulations:

• 0.3 reflectance target

• 500 m and 1500 m flying heights

• Effect of skylight and path

scattered radiance decreases

towards longer wavelengths and

with increasing flying altitude

1500 m

500 m

Total Radiance

Path Scattered (C)

Single Scattered Path Rad. (B)

Ground Reflected (A+B+D+E+F)

Direct reflected (A)

11

Atmospheric influences • Scatterig

• Rayleigh scattering • On particles much smaller than light wavelength (air molecules, aerosols).

• Symmetric in forward and backward directions.

• Power of scattered radiation proportional to with λ-4.

• Causes the blue color of the sky.

• Mie scattering • On particles of the size of the light wavelength, i.e. aerosols, dust particles, sea salt.

• Strongly in forwad direction.

• Varies slowly with wavelength

• Non selective scattering. On partilces greater than light wavelength, e.g. rain drops, snow. Varies neither with wavelength nor scattering angle.

• Gaseous absorption, mainly due to water vapour, carbon dioxide, oxygen

• Atmospheric influences are typically estimated by using radiative transfer codes, mainly Modtran 5 or 6S

12

Object properties – Reflectance Reflectance as the function

of the wavelength

13

• Reflectance ρ = dФr/dФi

• Ratio of radiant exitance (M

[Wm-2]) to the irradiance (E

[Wm-2])

• Law of energy conservation -

> ρ [0,1]

• Reflectance factor R =

dФr/dФid

• Ratio of radiant flux reflected

by a surfacte to that reflected

by an ideal and diffuse

(Labmbertian) surface

(similar irration and reflection

conditions )

Object properties – reflectance anisotropy

φi

i

Light

source

Observer

North

φr

r

),(

),,,(),,,(

ii

rriirrii

E

L

14

BRDF

L(θi,φi,θr,φr) Reflected radiance

E(θi,φi) Irradiance

• BRDF – Bidirectional reflectance distribution function

• Observed reflectance is dependent on viewing and illumination directions

• Influences are the largest in the solar principal plane: the plane containing the illumination source (sun), object, and observer

• With photogrammetric cameras FOVs are +/-30°: the anisotropy has to be taken into account

• Object specific

Sources of anisotropic reflectance • Specular reflectance

• Hot spot

• Sunglint reflectance on rough water surfaces

• Leaf/vegetation reflectance (volume scattering)

• Shadow-driven reflectance (surface scattering, gap driven)

• Microscopic backscattering on some surfaces (e.g. rocks)

15

Beisl and Woodhouse 2004

16

17

Central radiometric quality indicators of

images

• Spectral sensitivity: Which wavelengths are observed?

• Spectral resolution: How many channels? Width of the channels?

• Radiometric resolution: How many grey levels?

• Dynamic range: How dark and bright objects? • Signal-noise-ratio (SNR), Saturation

• Point spread function

• Grey value histogram

• Stability

• Calibration quality

• Traceability

18

Radiometric sensor calibration

19

Sensor/camera model (1) platform altitude

platform motion

at-sensor

radiance sensor imaging

optics

detectors electronic

s

A/D DN

spectral filters

or

dispersion

element

),(4

)(),(

2

0 yxLN

yxE i

max

min

),()(),(

dyxERyxs i

bb

ddyxPSFsyxe bb ),(,),( max

min min

max

)),(int( bbbpb offsetyxegainDN

),( yxL

Sensor model (2)

0

)()( dSLGADN sensoratd

DNf

KL

number

2'

The DN value of a given pixel, after dark pixel substraction

is applied:

Where

G system gain,

Ad area of detector,

Ω lens solid angle (aperture),

τ integration or exposure time,

S(λ) system level spectral response,

λ wavelength

Band averaged at-sensor radiance

21

Radiometric system

model: transformation

between the at-sensor

radiance and DN

Radiometric calibration - parameters • Relative pixel wise calibration: normalize output

of all detectors to the similar level and eliminate systmatic errors (noise in space)

• Photo Response Non-uniformity (PRNU): Sensitivity differences of individual CCD-elements

• Lens fall-off: Due to the drop of light intensity towards the edges of the focal plane

• cos4-law Eθ = E0 cosnθ

• Dark signal (DS), Dark signal Non-uniformity (DSNU)

• Uncorrected errors contribute to the signal noise

• Absolute calibration: relationship between the incoming radiance and DN

• Spectral response with respect to wavelength: central wavelength, badwidth

• Calibration methods: Laboratory, In-flight, Vicarious

DNf

KL

number

2'

22

Spectral response of photogrammetric

cameras DMC

0.0

0.2

0.4

0.6

0.8

1.0

350 450 550 650 750 850 950 1050

Wavelength

PAN

R

G

B

NIR

UltraCamD

0

4

8

12

16

20

350 450 550 650 750 850 950 1050

Wavelength

PAN

R

G

B

NIR

DMC II

23

ADS

Spectral response of some satellite

sensors and ADS

24

Laboratory calibration, ADS40 as example

• Absolute and relative: NIST traceable calibrated, homogeneous and isotropic light source illuminating the entire lens over the full aperture -> Integrating sphere

• Spectral: NIST tracable monochromator with a broad band light source and a spectrogrph to provide light with narrow spectral bandwidth

• Dark current: • Laboratory: black image

• Flight: Imaging with closed shutter, using black pixels in CCD

• NIST: National institute of Standards and Technology

U. Beisl, 2006: ADS 40

25

Why vicarious radiometric cal/val?

• ” The requirement for accurate radiometry is a thorough

understanding of the measurement problem, a complete

description and understanding of the instruments, and

mechanisms for comparing and assessing results “

(Johnson et al. 2004)

• Is laboratory calibration valid in flight conditions?

• What is quality of output products of an imaging system

(reflectance calibration etc.)?

• System cal/val after updates and in beginning of season

• Tuning image collection process, e.g. exposure

• Continuous change of systems

26 26

Photogrammetric process vs. imaging

system

Image product generation system

Referencing

Radiometric

correction

Georeferencing

Restoration

GIS

Image acquisition

Mission

design

Post-processing Image collection Mission

preparation

Measurement and

interpretation

Measurement

Interpretation

Calibration

Image acquistion system

27

Vicarious radiometric

cal/val

d1_g5: H = 500 m

0

500

1000

1500

2000

2500

3000

3500

4000

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18

At-sensor radiance

RA

W D

N

PAN

R

G

B

NIR

28

Requirements

•Accurate reflectance targes

•Reflectance characterization of targets

•Atmospheric observations

•Software for radiative transfer

Traceable Reflectance measurement chain

29

Laboratory at Metrology Research

Institute of Aalto University

MetEOC: Metrology for Earth Observation and Climate

CEOS test site catalog,

radiometry sites

30

Use of different calibration methods

Parameter Lab In-flight Test-field

Relative radiometry x x

Light falloff x x

Dark current x x

Spectral response x

Absolute radiometry (sensor) x x

Absolute radiometry (system) x

System validation x

31

Radiometric calibration of different

photogrammetric sensors • ADS40: NIST traceable calibration (Beisl 2006)

• Relative calibration: PRNU, Light falloff

• Dark current, DSNU

• Absolute calibration: Gain, offset

• Spectral response

• DMC : NIST traceable calibration (Pagnutti et al. 2009)

• Relative calibration: PRNU, Light falloff

• Dark current, DSNU

• Absolute calibration: Gain, offset

• Spectral response

• UltraCam (Calibration protocol)

• PRNU,defect pixels, light falloff

• Methods are still developing

32

Radiometric processing

33

Reflectance images by physical

correction method In typical photogrammetric imaging conditions, radiance entering

the sensor

Lat_sensor = Ls + Lsky + Latm

34

ρ=π(Lat-sensor - Latm)(1-sρ

)/(S τs τv)

s Spherical albedo of atmosphere, i.e. the fraction of the upward

radiance that is backscattered by the atmosphere

ρ

Average reflectance of the surrounding area

1-sρ

Term to take into account the multiple scattering

S Mean solar irradiance in surface S=Eλ0cosθs

τs τv Transmittance in solar and viewing paths. Derived using radiative

transfer models and image information

Reflectance

Reflectance images by Empirical line method

• Parameters are determined using 2 or more reference targets

• Artificial panels or natural surfaces measured by spectrometer

• Sometimes also spectral libraries are used to provide reference values of typical targets

• Dark vegetation is commonly used as dark reference

• Simple and functional method if reference targets are available in the campaign area.

• Commonly used in remote sensing

• In photogrammetric mapping projects this is not typically situation.

• Parameters are valid only for the specific campaign conditions

35

ρ = c1´ DN + c0´

Shadow correction

• Types of shadows: cast shadow and self shadow

• Shadowing object can be opaque or transparent, resulting in full

(uniform, e.g. building) shadow or partial (non-uniform, e.g. tree)

shadow.

• No definite boundary between shadowed and non-shadowed

regions.

• Cloud shadows

• Shadow detection

• Using a 3D surface model

• Using image information: e.g. ATCOR-ADS : signal in blue spectral

band is much higher (Schläpfer 2012)

36

Additional corrections

• BRDF-correction: Correction of directional reflectance

effects to obtain uniform image mosaics etc.

• Class specific BRDF-correction needed

• Changing illumination conditions

• Adjacency effects

• Topographic correction

37

ATCOR4: Atmospheric correction of airborne

multi- and hyperspectral images • ATCOR-4: Atmospheric & Topographic Correction for

wide FOV airborne optical scanner data

• Based on large “monochromatic” atmospheric database for altitudes between 1-20km, compiled with the MODTRAN5 –code.

• Basic output is surface reflectance cube i.e. image with the original image-DN’s converted to surface reflectances. Options

• Correction of topographic effects

• Empirical, class-specific BRDF correction

• De-shadowing of cloud/and other shadow areas

• Statistical haze removal

• In-flight sensor calibration using ground reference targets

• Algorithm for automatic classification of reflectance spectra

• Calculation of value adding products: SAVI, LAI, FPAR, albedo, surface energy fluxes

38

ATCOR4: Requirements for

atmospheric correction • Accurate radiometric calibration of the sensor

• Either from the sensor manufacturer or by performing in-flight calibration using 1, 2 or >2 ground reference reflectance targets

• An accurate estimation of the main atmospheric parameters • Aerosol type: rural, urban, maritime, desert

• Can be estimated in ATCOR based on images, if image contains dark vegetation

• Visibility (or aerosol optical thickness)

• User given, interactive estimation or automatic calculation (requires red/NIR –channels and dark targets)

• Water vapor

• Automatic calculation or estimation based on the season

39

Leica ADS40 and XPro • The first reflectance calibration in photogrammetric post-processing software.

• Correction is based on high quality, calibrated sensor, physical/empirical

atmospheric modeling and image data.

• Fulfils the usability requirements of mapping agencies: fully automatic,

ground reference targets are not needed.

• Prosessing in XPro

• At-sensor radiance by laboratory calibration DN -> Lat-sensor

• Directional reflectance

• Atmospheric reflectance is based on dark pixel method. Unknowns : L0,τup,

τdown and s are solved by regression based on simulated atmospheric data.

The explicit form of the regression functions have not been published.

• BRDF-correction to eliminate object anisotropy

• Similar method has been implemented for DMC and RCD30 frame

images, available soon

)1()( 0

s

S

LL

downup

sensorat

40

Image from U. Beisl

Beisl, U., Telaar, J., Schönemark, M. v., 2008. Atmospheric correction,

reflectance calibration and BRDF correction for ADS40 image data.

ISPRS Peking Symposium, VII, WG VII/1.

41

Radiometric block adjustment, Pepita

of IGN France

• A method for equalising a block of digital aerial images.

• Can be considered as relative radiometric calibration.

• Based on a parametric, semi-empirical radiometric model taking into account BRDF, haze differences between blocks of images, solar elevation, sensor settings (exposure) and chromatic aberrations.

• Does not require calibrated sensor

• The model parameters are computed through a global least-squares minimisation process, using radiometric tie-points in overlapping areas between the images.

• The method has been used in the IGN/France orthoimage workflow since 2005. Fully operational.

• Pepita type method has been implemented in UltraCam post-processing software UltraMap

42

IGN/France Pepita model

• Unknowns

• For each image: k, e, a

• For each image block collected in similar conditiosn: Lhaze

• Radiometric tie points are collected over the block area.

Minimization task

43

Solar

irradiance

BRDF-model

Radial term to

correct chromatic

aberration

Exposure

time

Hazeterm Visibility and

solar elevation

Ground

reflectance

Observed

at-sensor

radiance

Chandelier, L.; Martinoty, G. Radiometric aerial

triangulation for the equalization of digital aerial

images and orthoimages. Photogramm. Eng. Remote

Sens. 2009, 75, 193-200.

FGI’s radiometric block

adjustment for frame images

• Data • Overlapping spectral rectangle format data cubes

• Tasks • Eliminate radiometric disturbances caused by sensor instability and

illumination/atmosphere

• BRDF compensation

• Reflectance calibration

• Approach • Radiometric model parameters using radiometric block adustment with a

network of radiometric tie points

• Optional insitu irradiance measurements

• Reflectance images using reflectance targets

44

F

D A A

B

C E G

O

φi

i

Light

source

Observer

North

φr

r

Radiometric end-products of digital

photogrammetric cameras • Radiometric end products: Reflectance image, true color image,

relatively radiometrically corrected image, something else • ADS40: Chain for reflectance image generation

• NIST traceable laboratory calibration • Xpro: processing of line scanner images

• Atmospheric correction: path radiance, adjacency effect • Reflectance calibration • Semi empirical BRDF correction

• DMC: • NIST traceable laboratory calibration • PPS: Radiometric correction of frame images, similar to Xpro (soon available!)

• UltraCamD • Laboratory calibration: relative • Model based radiometric equalization, similar to Pepita

• New, independent approaches are under development: • ATCOR-4 for ADS • ICEE • Pepita of IGN • FGI radiometric block adjustment • …

45

EuroSDR project

46

47

EuroSDR project: ”Radiometric aspects of

digital photogrammetric airborne images”

• Schedule: 2008-2011

• Objectives • Improve knowledge on radiometric aspects of digital photogrammetric

cameras

• Review existing methods and procedures for radiometric image improvements

• Compare and share operative solutions through a comparison of these techniques on a same test data set

• Analyse the benefit of radiometric calibration in order to open new applications (classification, quantitative remote sensing, change detection etc.)

• Headed by • FGI, Finland: Eija Honkavaara, Lauri Markelin

• ICC, Spain: Roman Arbiol

47

EuroSDR project

• Phases

1.Review

2.Empirical research

• Participants

• European National Mapping Agencies (6)

• Software development (1)

• Universities and research organizations (8)

• Many different points of view are covered

48

EuroSDR Review: Major conclusions

• Improvements are requested for the entire process: sensors,

calibration, data collection, data post-processing and data utilization.

• Fundamental problems:

• Insufficient information of radiometric processing chain

• Inadequate radiometric processing lines

• Missing standards (methods, calibration, targets, terminology)

• The basic radiometric end products requested by image users: true

color and reflectance images.

• Expected benefit of more accurate radiometric processing:

• more automatic and efficient imagery post-processing

• better visual image quality

• more accurate, automatic interpretation , remote sensing use

• Only one participant had implemented theoretically based radiometric

correction in 2009. Situation changing since 2011

49

Stakeholders

Type Sensor manufacturing Software

development Data collection Image products Applications Research

U1 x x x x x (x)

U2 x x x x (x)

U3 x x x (x)

U4 x x (x)

U5 x (x)

P1 x x x (x)

P2 x x x x (x)

P3 x (x)

P4 x x (x)

R1 x

SW1 x (x)

M1 x x (x)

Data users perspective:

•Different: sensors, calibrations, processes, settings…

•Changing: sensors, calibrations, processes, settings…

•Subjective selections, non-transparent process….

Sensor Software Applications Image

products

Data

Collection

50

Phase 2 – Materials • Comprehensive campaigns in 2008 with

atmospheric observations and reflectance targets

• DMC + CASI: Banyoles Spain • Flying heights: 820, 1125, 2250 and 4500 m.

• DMC: Sjökulla, Finland • Flying heights: 500 m

• ADS40: Hyytiälä, Finland • Flying heights: 1000, 2000, 3000 and 4000 m • Forest test site

51

Reference targets, Hyytiälä Finland

4 tarps, sand, gravel, weeds, grass

Reflectance factors by ASD Field

Spec Pro FR + 12’’ Spectralon

52

53

Evaluated radiometric correction

methods

• Reflectance images

• Leica Geosystems Xpro

• ReSe ATCOR-4

• Empirical line

• Relative block adjustment

• IGN Pepita

54

Accuracy assessment

• Reflectance error in reflectance units

Erefl=ρdata – ρref

• Reflectance error in %

Erefl%=100(ρdata – ρref)/ ρref

• RMSE

RMSErefl% = √(∑(Erefl%2 )/n)

• Anisotropy scaling

ρdata_nadir = ρdata ρlab_nadir/ρlab_exact

55

Reflectance spectrums of

asphalt and grass by XPro

0

0.05

0.1

0.15

0.2

0.25

400 500 600 700 800 900

Re

fle

cta

nce

Wavelength [nm]

Asphalt

10cm

20cm

30cm

40cm

Ref

0

0.1

0.2

0.3

0.4

0.5

400 500 600 700 800 900R

efl

ect

an

ce

Wavelength [nm]

Grass1

10cm

20cm

30cm

40cm

Ref

56

Markelin, L., Honkavaara, E., Schläpfer, D., Bovet, S., Korpela, I., 2012.

Assessment of radiometric correction methods for ADS40 imagery. .

Photogrammetrie, Fernerkundung, Geoinformation (PFG) 4/2012

Reflectance RMSEs on bright

tarps • Xpro

• Sensor: Lab calibration

• Atmospheric parameters from images

• ATCOR-4

• Sensor: 1) Lab calibration, 2) Self-calibration (reference targets in image block), 3) In-situ calibration: using single flight line in 1 km to calibrate sensor

• Atmospheric parameters from images

• Best results with Xpro and ATCOR-4 with self calibration

• RMSE on level of 5% at best

• Accuracy influenced by flying height, atmosphere, channel, magnitude of reflectance.

57

Results of Xpro Correction

• plaa

58

At-sensor radiance At-sensor radiance + BRDF

Atmospheric correction Atmospheric correction

+ BRDF

Downey. M., Uebbing, R., Gehrke, S., Beisl, U., 2010. Radiometric processing of ADS

imagery: Using atmospheric and BRDF corrections in production.ASPRS 2010 Annual

Conference, San Diego, California. April 26-30, 2010.

59

Results of Pepita of IGN

• The solar elevation, exposure time and aperture values were well

corrected.

• The BRDF model worked well especially on forest and agricultural parcels.

• Potential problems

• BRDF model did not fit on the lake, high residuals

• In urban areas: the radiometric model may not be convergent

Chandelier, L., Martinoty, G, 2009. Radiometric aerial triangulation for the

equalization of digital aerial images and orthoimages. Photogrammetric

Engineering & Remote Sensing 2009, 75, 193-200.

60

Example 1. Tree species

classification using reflectance

calibrated photogrammetric images • Based on

• Heikkinen, V., Korpela, I., Tokola, T., Honkavaara, E., Parkkinen, J., 2011. An SVM classification of tree species

radiometric signatures based on the Leica ADS40 sensor. IEEE Transac¬tions on Geoscience and Remote

Sensing 49(11), 4539 – 4551.

• Korpela, I., Heikkinen, V., Honkavaara, E., Rohrbach, F., Tokola, T., 2011. Variation and directional anisotropy of

reflectance at the crown scale - Implications for tree species classification in digital aerial images. Remote

Sensing of Environment 115 (8), 2062-2074.

61

Research questions

• Development of methods for tree species classification

using high resolution photogrammetric images

• Reflectance characteristics of tree species: Scots pine,

Norway spruce and birch

• Influence of image data processing level on classification

results: at-sensor radiances, directional reflectance (BRF),

nadir normalized reflectance (BRDF corrected)

• Quadratic discriminant analysis and support vector

machine classification

• Generalization properties of reflectance factors

concerning the need for in-situ training data

62

Determination of visibility and

shading conditions • Estimation of crown

envelope with dense

lidar data: 121 points

per cown

• Illumination condition

for each point by ray

tracing:

• Occlusion: Cloud, tree

itself, adjacent tree

• Shadow: Cloud, self-

shaded, neighbour

shaded, neighbour and

self-shaded

63

© Ilkka Korpela 2011

Illumination in forest

64

© Ilkka Korpela 2011

Reflectance characteristic of tree species

• Plottings as the function of FOV-angle against to the flying direction, approximately in solar principal plane

• Near the solar principal plane, pine, spruce and birch showed anisotropy of ± 30%, with differences between illumination classes, bands and species.

• The within-species variation was high; the coefficients of variation were

13-31% when normalized to the reflectance at nadir.

65

At-sensor radiance Atmospheric

compensation

Atmospheric

compensation,

BRDF correction

Classification results

• Classification experiments with reflectance calibrated data

provided an accuracy of about 75-79% with single ADS40

view and 78-82% with two ADS40 views (Heikkinen et al.,

2011).

• In some cases (the training and classification data from

different flight lines) the use of reflectance calibrated data

provided better classification results than the at-sensor

radiance data

• Empirical BRDF correction did not improve classification

results

66

Example 2. NDVI time series

over the Catalonia by ICC

© Roman Arbiol, Lucas Martínez

67

Process

68

LR4 Images from DMC (without post-process)

Absolute radiometric calibration (by manufacturer or by ICC)

Photogrammetric direct block orientation

Public free WMS service (http access also with OrtoXpres)

(http://www.ortoxpres.cat)

• OrtoXpres fast publication after the flight (a few weeks)

© Roman Arbiol, Lucas Martínez

69

© Roman Arbiol, Lucas Martínez

70

© Roman Arbiol, Lucas Martínez

NDVI WMS layer

71

NDVI WMS layer

1st NDVI layer in 2011. Catalonia (32,000km2 & 25cm GSD)

8bpp (DN from 0 to 255)

Orto-rectified “on the fly” of single photographs without stiching

(all zoom levels are not available)

Service available for most of the SIG environtments : ArcView,

Miramon, gvSIG, etc.

NDVI layer users

Agriculture Department of the Generalitat de Catalunya (Catalonia

regional government) will use NDVI layer it to verify agriculture

policy (ie. wineyard, cereal, etc.)

After, the vegetation layer will be freely disseminated according to

ICC data policy

© Roman Arbiol, Lucas Martínez

Example 3. Influence of solar

elevation in photogrammetric

processes and digital image matching

• Honkavaara, E., Markelin, L., Rosnell, T., Nurminen, K, 2012.. Influence of

solar elevation in radiometric and geometric performance of multispectral

photogrammetry. ISPRS Journal of Photogrammetry and Remote Sensing.

67, 13-26.

72

Flight strips in Hyytiälä test site

• Image collection using flying heights 2, 3 and 4 km, GSDs 20, 30, 40 cm, by National Land Survey of Finland DMC

• Data processing in Bae systems Socet Set 5.5 environment, DSMs by NGATE

• Solar elevations: Morning: 25-28º, Noon: 44-48º

73

General evaluation: Strip histograms

• Exposure and aperture were tuned to provide desired histograms

• Strip histograms were not dependent on solar elevation

• Histograms were similar in morning and afternoon

Morning

Noon

2 km 4 km

74

DSM extraction in deep shadows (1)

Height

artefacts in

shadows

75

DSM extraction in deep shadows (2)

• Height variation 2-3 times higher in

shadows

Noon Morning

0

0.1

0.2

0.3

0.4

0.5

0.6

20 30 40 20 30 40 20 30 40 20 30 40

morning noon morning noon

shadow field sunny field

av_sdev

76

Results

• Histograms similar in morning and noon

• Point cloud point density

• Typically the point interval of GSD successful

• Detailed analysis showed that in sun illuminated areas point

cloud quality similar

• Potential problems in shadowed areas

• In planar targets increased height variation

• Failures of measuring road surfaces in deep shadows

• Reflectance calibration difficult

• Low dynamic range

77

Example 4. Radiometric processing in

hyperspectral UAV remote sensing

78

UAV operation

ISPRS Hannover Workshop 2013, 21 – 24 May 2013,

Hannover, Germany 79

UAV

•Autopilot

•IMU

•GPS

Payload

•Spectral imager

•High spatial

resolution imager

•GPS

•Irradiance sensors

Ground control station

•Mission design and control

•Insitu reference measurements: irradiance,

reflectance targets,

•In typical flight 100-500

data cubes with 20-40

spectral layers

•Georeferencing data

•Irradiance data

•Insitu data

Fabry-Perot interferometer based

tuneable spectral camera • Hyperspectral imagery in frame

format: stereoscopic, spectrometric data

• Weigh 600 g

• Spectral data cube by changing the width of Fabry-Perot air gap

• Developed by VTT Technical Research Finland (Heikki Saari)

• Custom optics: C=10.9 mm, F-number < 3.0

• CMOS detector: 1024 x 648 pixels, Pixel 11 μm

• Application based filter selection between 400-1000 nm

• 500-900, 600-1000, 400-500, … nm

• Spectral resolution 10-40 nm @ FWHM

• Commercial systems by Rikola Oy

ISPRS Hannover Workshop 2013, 21 – 24 May

2013, Hannover, Germany 80

Vihti campaigns

81

• MTT agricultural test area in Vihti

• Hyperspectral UAV camera

• July 2, 2012 10:39 and 10:50 local time (UTC +3).

• Poor illumination conditions with fluctuating levels of cloudiness

• Flying altitude 140 m -> GSD of 14 cm

• Irradiance measurements by UAV and in ground

• Radiometric block adjustment

0

0.5

1

1.5

2

2.5

3

Image

ground, 29

uav

BA: relA, 29

average, 29

Correction factors to reference image

Image mosaics and sample spectra, FGI

82

No corr

UAV

Ground

Image

based

Coefficient of variation

• Average

coefficient of

variation in tie

points

• Without correction

0.14-0.18

• UAV: 0.1-0.12

• Ground: 0.06-0.09

• Block adjustment:

0.04-0.07

83

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

500 600 700 800 900

Wavelength (nm)

no corr

uav

ground

BA: relA

Coefficient of variation

Biomass estimation by knn-estimator

84

No correction

R2: 0.64

NRMSE: 24.9%

Ground irradiance

R2: 0.74

NRMSE: 17.8%

Block adjustment

R2: 0.74

NRMSE: 16.8%

Concusions

85

Data processing • Accurate and efficient radiometric processing methods are now available

• Accurate calibration methods required • Laboratory • Test field

• Reflectance images by utilizing sensor calibration, atmospheric models and BRDF correction

• With 1 st generation methods, 5% accuracy was demonstrated

• Operational: no ground control needed, fully automated

• Solutions for all major photogrammetric cameras: ADS, DMC, UltraCam

• Improvements still needed to the methods • Shadowed areas

• Topographic correction

• Class specific BRDF correction

• Absolute reflectance

• Temporal differences

• NMAs are implementing next generation radiometric correction methods in their processing lines

• Swisstopo: Swissimage standard, Remote sensing basic

• IGN France: Pepita

• ICC

• …

86

Outlook • Mutlispectral photogrammetric imagery are suitable for

remote sensing use.

• Lots of unused potential.

• What is the most cost efficient approach for large area

accurate topographic mapping, forest information

derivation, long time GIS data maintenance, etc.?

• Reflectance calibrated multiple directional observations

• BRDF corrected normalized reflectance

• Campaign specific classifications without controlled radiometry

• Single sensor vs multiple sensor applications

• Feasiblity of 3D reconstruction and ray tracing in object

interpretation?

• Many advantages are expected from rigorous processing.

87

Thank you!

Questions?

88

top related