Lecture 6

Photogrammetric practices

Photogrammetric practices
• Photogrammetric measuring processes consist of four subsequent
phases, in which the 3D objects and scenes are interpreted from
images and geometrically transformed to a 3D model.
• The phases are
• image acquisition,
• image preprocessing,
• interpretation of images in 3D
• presentation of objects and scenery as a 3D model.

Reading:
Tony Schenk, Introduction to Photogrammetry, 2005 http://www.mat.uc.pt/~gil/downloads/IntroPhoto.pdf

1
Photogrammetric process

Image acquisition

2
Aerial photography
Stereokuvaus

Laserkeilaus

(Uwe Lohr, TopoSys)

Digital photogrammetric cameras

(Eija Honkavaara, 2014. Maa-57.3100 Käytännön fotogrammetria)

3
 Hexacon airborne imaging systems

Leica ADS
Pushbroom camera

Leica ALS
Laser scanner
LIDAR
Leica RCD
Medium format
frame camera

Z/I DMC II Large format frame camera

(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

Pushbroom vs. frame imaging

(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

4
Leica ADS80 - CCD layout in focal plane

(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

Leica ADS80

(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

5
Leica ADS80 Airborne Digital Sensor
• Multispectral Pan and RGBN sensor
• Innovative beamsplitter design provides equal resolution in all bands
• 12000 pixels, 6.5um, in all bands in standard mode (Ratio 1:1)
• 24000 pixel swath in HiRes Mode (Ratio 1:2)
• Focal length 65mm (single lens design)
• 100% forward overlap at all times
• Continously recordable strip length
• Dedicated Workflow
• Very Rapid orthoprocessing and AT
• High accuracy for DSM extraction
(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

Z/I DMC II

(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

6
Z/I DMC II
~ 90 mm

(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

Z/I DMC II

(Rohrbach, 2012. https://www.terrasolid.com/download/presentations/2012/hexagon.pdf)

7
DMC and DMC II Technical Specifications

(Neumann, 2012)
http://calval.cr.usgs.gov/wordpress/wp-content/uploads/Neumann_ZI_DMC_II_JACIE_2012.pdf

Microsoft Vexcel Ultracam

(Eija Honkavaara, 2014. Maa-57.3100 Käytännön fotogrammetria)

8
UltraCam Osprey • Mark 3 Premium

Aerial photography (National Land Survey)

• Ground control measurements
• Photography and laser scanning
• Post processing
• Ground control coordinates
• Exterior orientations for images with
reference to GNSS frame
• Preliminary processed laser data with
reference to GNSS frame
• Quality control
• Time series
• Nearly 1 million images since 1930

9
Intergraph Z/I DMC

(Eija Honkavaara, 2014. Maa-57.3100 Käytännön fotogrammetria)

NLS Register of aerial photography

http://hkp.maanmittauslaitos.fi/hkp/published/fi/4343c1b4-7d8f-4473-896a-70f930f36be1

10
NLS File service of open data

https://tiedostopalvelu.maanmittauslaitos.fi/tp/kartta?lang=en

100 000 50 000 25 000 10 000 5 000

48 km x 96 km 24 km x 48 km 12 km x 24 km 6 km x 6 km 3 km x 3 km

11
 Aerial photography

b = image base
c = camera constant (focal length)

DMC field of view

120 mm 120 mm

69,4° 42°

12
 Sensor size 166 mm

DMC specifications
Focal length 120 mm 92mm
Image size 13824 pixels x 7640 pixels
Pixel size 12 μm x 12 μm
Field of view 69,4° x 42 °
12 μm

120 mm Example:

In case ground sample distance GSD is 50 cm,

FH?
Flying height FH?
50
= 120 = 5000
12 μ

GSD 50 cm

Flight planning

58 km

40 km

13
 Image size 13824 pixels x 7680 pixels
Image block Ground coverage 6912 m x 3840 m

Overlap 60%
Stereo pair (Stereo model) Stereo model coverage 2304 m x 6912
• 60 % of 3840 m
Forward overlap 60 %
• 100 % of 6912 m

Strip formation:
Image base 1536 m
• 40 % of 3840 m
First image taken at 3840 m /2 = 1920 m
Strip 2nd at 1920 m + 1536 m
3rd at 1920 m + 2x1536 m
etc.

Side overlap 20-30 % Block formation:

Strip interval 4838 m
• 70 % of 6912 m
First strip at 6912 m /2 = 3456 m
2nd at 3456 m + 4838 m
3rd at 3456 m + 2x4838 m

Image block
Stereo pair (Stereo model)
Forward overlap 60 %

Strip Block

Side overlap 20-30 %

14
 Image blocks

(Kimmo Nurminen)

https://www1.ethz.ch/igp/photogrammetry/education/lehrveranstaltungen/PhotogrammetryH
S12/06-E-flightplanning.pdf (Enni Airaksinen, Kari Kaisla, Jarmo Suomisto, 1.9.2016)

GNSS Global navigation satellite system

GPS, GLONASS, Galileo or Beidou

15
 IMU Inertial measurement unit

(King, A.D., GEC Review, Vol. 13, No. 3, 1998)

GNSS/IMU synergy
GNSS IMU
+ Good long-term accuracy + High short-term accuracy
– Low frequency (1-2 Hz) + Accurate rotation information
– Cycle slips, multipath + High frequency
– Poor rotation accuracy – Poor long-term accuracy

GNSS/IMU
+ Positions and rotations
+ Good long-term and short-term accuracy
+ High frequency

16
GNSS/IMU for preliminary image orientations
• Airplane
• GNSS-receiver
• IMU (Inertial Measurement Unit)
• PC + navigation software
• Ground
• In most cases 1 or more GNSS base stations <50 km distance or VRS Virtual
Reference Station network or corresponding.
• Post processing software
• Carrier phase differential GNSS solution (typically)
• GNSS/IMU integration by Kalman filtering

X0 Y0 Z0 , and omega-, phi- and kappa-rotations for every image

Post-processing of GNSS data

http://www.novatel.com/an-introduction-to-gnss/chapter-5-resolving-errors/gnss-data-post-processing/

17
UAV blocks

(Markus Gerke, 2012)

Preprocessing

18
Block adjustment
Image observations

Image orientations

Object coordinate system

Object coordinates for tie points

Block adjustment and aerial triangulation

• Aerial triangulation is the most accurate photogrammetric task in mapping
process
• Point densification (tie points)
• Based on geodetic networks (control points)
• By aerial triangulation we produce coordinates for
• Land marks and border points
• Natural features which are visible in the photos to be used as control points for stereo
restitution and orthophoto production

Control point Tie points

19
 Corrections for collinearity ③
Principal point Optical distortion
①
Camera frame geometry ②
Corrections to ′ -coordinates

Corrections to -coordinates
Projection centre
Refraction
1 −
③ Collinear image ray ′ = −
− −
Map projection
①

(Ref: e.g. Schenk, 2005)

② Geoid http://www.mat.uc.pt/~gil/downloads/IntroPhoto.pdf

GNSS/IMU for aerial triangulation

20
Tie point selection
Image no. Tie points Tie points
Tie points image 8 image 11
1 1
2 4
3 7
4 4
5 1
6 3 7
7 10 4
8 18 1
9 10
10 3
11 1 17
12 4 10
13 7 3
14 4
Control points 10 1

Image residuals (dx, dy)

Residual (dx,dy)Image 1
P

Residual (dx,dy)Image 3
Residual (dx,dy)Image 2

21
 Tie points for stereo restitution
• A minimum requirement is,
that for each stereo model
we must observe at least
three such control points,
which are known or
observed in the object
coordinate system as well.
• Practically there are always
more control points and they
cover widely the entire model
area.

Scale, shift and rotations vv

v
v

22
 Absolute orientation
2. Sequentially, by
1. Directly with GPS-control points, or … • Scaling
• Levelling
• Rotation, and
• Shift

(Markus Gerke, 2012)

Analysis

23
EspaSystems ESPA Systems makes innovative software for digital
photogrammetric mapping and LIDAR data processing. Our
products are used by national land surveys, city mapping
offices, mapping and civil engineering companies, forestry
organisations and geological surveys.

Maanmittauslaitoksen ylläpitämän valtakunnallisen maastotietokannan tiedonkeruussa

on vuodesta 2001 käytetty EspaCity-sovellusta.
Työasemalla tulkitaan digitaalisia stereo-ortokuvia ja maastoa editoidaan
suoraan Smallworldin maastotietokantaan.
Maastopohja kerätään laserkeilatusta LIDAR-aineistosta.

Image orientation

24
Stereo restitution

(Hannu Hyyppä, Petri Rönnholm, TKK)

25
Intersection in space (stereo images)
Repeatability ≈ Precision ( dφ )

C = 120 mm
B = 1540 m

FH = 5000 m

Point measuring in stereo photogrammetry

Aerial photography in national
RMSEX = RMSEY ≈ 0,5 GSD topographic mapping:

GSD = 50 cm
c = 120 mm
FH = 5000 m
RMSEZ ≈ 1 GSD B = 1540 m

Accuracy of aerial triangulation:

dx = dy = 0,005 mm
dpx = 0,0035 mm

26
3D measuring accuracy of stereo models
dX, dY =

=
dZ 2

dx, dy Dpx
Stereoscopic parallax measurement

The accuracy of point observation on image can be projected to object space

Inaccuracy of distance observation

Scale: Base-to- Parallax:

height ratio:

27
Example
Aerial photography in national
topographic mapping:

GSD = 50 cm
c = 120 mm
FH = 5000 m dX = dY = 21 cm
B = 1540 m dZ = 48 cm

Accuracy of aerial triangulation:

dx = dy = 0,005 mm
dpx = 0,0035 mm

Measurement accuracy
• Accuracy of observed image points
• with correct GCP pattern and correct processung, the influence factor is the
image scale
• Measurement accuracy depends on
• definition uncertainty of points
• image quality
• lighting conditions
• automatic/manual measurement
• Empirical accuracy
• for well-defined points 0.3-0.7 Pixel, depending on conditions
• Note: often clients specify a maximal allowed inaccuracy, calculate with 3σ ≈ 1.0-
2.0 Pixel
• relevant points often have high definition and interpretation uncertainty

https://www1.ethz.ch/igp/photogrammetry/education/lehrveranstaltungen/PhotogrammetryHS12/06-E-flightplanning.pdf

28
 Esimerkki stereomittauksesta

Kuva: Petri Rönnholm

57

Scene presentation

29
NLS Digital elevation data 32 m, 8 m, 2 m
Stereo restitution
10 m x 10 m Laser scanning
≥ 0,5 pts/m2

Orthophotography
• Coordinate systems
• Aerial photograph
• Map
• Image functions
• Aerial image
• Orthophoto
• True orthophoto
• Ortho mosaick
• Stereokuva
• Stereoscopic image pair
• Stereo-orthophoto

30
Aerial photo

Orthophoto

31
True orthophoto

Collinear perspective rectification

• Interpolation of Z from DEM ( x, y ) = f ( X , Y , Z )
• Calculation of image
coordinates
• Interpolation of radiometric
function

32
 Interpolation of DEM

Z = f ( X ,Y )

Bilinear interpolation within DEM

A 0 0 7
B 1 0 9
C 1 1 8
D 0 1 6
E 0,6 0,4 7,8

33
Interpolation on image

A 0 0 7
B 1 0 9
C 1 1 8
D 0 1 6
E 0,6 0,4

Color and Color Infrared orthophoto

34
3D modelling

(Katri Koistinen, TKK)

(Arzu Çöltekin, TKK)

1 : 10 000 000 10 km *
1 : 5 000 000 5 km G
1 : 2 000 000 2 km L
1 : 1 000 000 1 km O *
1 : 500 000 500 m B N
1 : 200 000 200 m A A
1 : 100 000 100 m L T
1 : 50 000 50 m * I *
1 : 20 000 20 m O R
1 : 10 000 10 m N E
1 : 5 000 5m A G
1 : 2 000 2m L I
1 : 1 000 1m * O *
1 : 500 50 cm N L
1 : 200 20 cm A O *
1 : 100 10 cm L C I
1 : 50 5 cm * A N *
1 : 20 2 cm L D P
1 : 10 1 cm * O R
1:5 5 mm O O
1:2 2 mm R D
1:1 1 mm * U
2:1 0.5 mm C
5:1 0.2 mm T
10 : 1 0.1 mm *

35
(Enni Airaksinen, Kari Kaisla, Jarmo Suomisto, 1.9.2016)

(Enni Airaksinen, Kari Kaisla, Jarmo Suomisto, 1.9.2016)

36
 (Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

Helsinki3D+
• Photography
• Helsinki, 400 km2
• GNSS/INS
• Aerial photography
• Texturing, mesh model, oblique images
• Microsoft Osprey multi-frame camera
• Ortho, nadir (RGB, CIR), 4 oblique
• Flying height 12370 m
• GSD 10 cm
• Side lap 65%, forward lap 80%
• Laser scanning
• 3D city model
• 20 pts/m2
• Flying height 475m / 530 m
• Side lap 50%, angle of view 35°

(Enni Airaksinen, Kari Kaisla, Jarmo Suomisto, 1.9.2016)

37
 Aerial photography and laser scanning
Aerial photography with UltraCam Osprey Prime (Vexcel) Laser scanning with ALS-70 HP and AKLS-70 CM (Leica)
• Flying height 1370 m • 530 m (HP) and 475 m (CM)
• GSD 10 – 15 cm • 20 pts/m2
• Over 50 000 images, 24 bit

Production of a mesh model

1. Aerial photography
2. Preprocessing
1. Tiling
2. Extraction of water areas
3. Control point measurements
3. Production
1. Camera and control data
2. Aerial triangulation
3. Model reconstruction

(Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

38
 (Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

Automatic aerial triangulation

(Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

39
 (Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

Aerial triangulation by ContextCapture

Residuals

Before

After

Tie points TIN DSM TIN + Textures

(TIN = triangulated
irregular network)

(Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

40
 Mesh model
(DSM + Textures)

Close-up details

(Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

Production of a city information model

Production of a city information model
1. Data acquisition
1. Airborne laser scanning
2. Aerial photography
2. Preprocessing
1. Processing of topography: DEM and DSM
2. Ortophotography and extraction of building outlines
3. Production of CityGML-objects
1. Geometric and semantic processing of buildings
2. Automatic texturing of building surfaces
4. Production of the DEM and ortophotography

41
(Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

(Open Geospatial Consortium, Inc., 2012)

42
DTM and DSM

(Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

CityGML LOD 2 + DEM + Phototextures

(Enni Airaksinen, 2017. Three-dimensional city modelling methods, Master’s Thesis)

43
LOD 1 + DEM LOD 2 + DEM LOD 2 + DEM + Textures

(Enni Airaksinen, Kari Kaisla, Jarmo Suomisto, 1.9.2016)

44
NLS Topographic database operators
Subcontractors Subcontractor Subcontractor

Producers Producer Producer Producer Producer

NLS coordinator NLS Topographic database

Aerial and Aerial Elevation Topographic

Client
orthoimages point cloud models objects

45