damen2010 geometric operations

7/28/2019 Damen2010 Geometric Operations

1/38

1

ra nng ourse

Remote Sensing Basic Theory & Image Processing

Methods

Michiel DamenMichiel Damen (September 2011)(September 2011)@@

ITC FACULTY OF GEO-INFORMATION SCIENCE AND EARTH OBSERVATION, UNIV. OF TWENTE


2/38

2

Geometric Operations

Contents

Two dimensional approaches

Georeferencing eoco n

Three dimensional approaches

Orientation

Stereo restitution

Questions

INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATIONDamen 2009

ITC FACULTY OF GEO-INFORMATION SCIENCE AND EARTH OBSERVATION, UNIV. OF TWENTE 2


3/38

3


Why correct images for geometric

distortions ?

INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATIONDamen 2009

ITC FACULTY OF GEO-INFORMATION SCIENCE AND EARTH OBSERVATION, UNIV. OF TWENTE 3


4/38

4


Why correct images for geometric

distortions ?

1. To get 2-dimensional (x, y) and 3 dimensional (x,

,

2. To visualize the image data in a GIS environment

3. To merge (fuse) different types of image data,for instance in a GIS or multi-tem oral anal sis

INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATIONDamen 2009ITC FACULTY OF GEO-INFORMATION SCIENCE AND EARTH OBSERVATION, UNIV. OF TWENTE 4

(monitoring)


5/38

5


Elementary image distortions

-

of geometric image



6/38

6



images of large scaler

H



7/38

7



Enschede



8/38

8


To be used if relief displacement can be neglected:

Two dimensional approaches

(1) flat terrain or (2) space imagery with low resolution

The objective is to relate the image coordinate system to

a spec cmap

coor nate systemImage Map



9/38

9


Georeferencing

1. Selection of the appropriate type oftransformation

.

A geometric transformation is a function that

for instance map x, y to image: i, j

The t e of transformation to be used de ends mainl


on the type of sensor-platform


10/38

10


Types of transformation

rans orma on : correc s or p c an

roll in for instance aerial photographs

st nd th, .

In most situations 1st order adequate

x = a + bi + cj x, y : map coordinates

y = d + ei + fj i,j : image coordinates

The six transformation parameters ( a. f) can be


determined by Ground Control Points ( GCPs )


11/38

11


Ground Control Points ( GCPs )

s : po n s acan be clearly identified both in the image and

in a source with the required map projection.

Sources: (1) topographical map, (2) alreadyrectified image, (3) GPS field points !

Master Slave Image Map



12/38

12


Rectification in ERDAS Imagine (image-to-image)



13/38

13


Accuracy of the transformation

: ca cu a es amean value from the individual residuals, both in x

and y direction.

! Warning !

The RMS error does not take into account the spatial

distribution of the GCPs and also not the accuracy of

the person trying to find the Control Points !!



14/38

14


Accuracy of the transformation

n

i

ix dxn

m1

n21

22)( yxtotal mmRMSEm

Root Mean Square Error


i

iyn

m1


15/38

15


Images Original and geo-referenced

Original image Geo-referenced image


Flevoland - IJssel delta, The Netherlands Co-ordinates


16/38

16


Geocoding -> Resampling

s geo-re erenc ng w su sequen

Resampling of the raster image.

1. Projection of raster onto

the original image using the

trans ormat on parameters

2. Resamplingof raster image



17/38

17


Transformation types for Geo-coding

,

based on their increase in complexity:



18/38

18



Conformal (4 parameters):

Rotation, scaling,

preservation of angles (shape)



19/38

19



Affine (8 parameters):

Rotation, scaling,

preservation of parallels



20/38

20



Polynomial (2nd order):

Rotation, scaling,

no preservation of parallelsstrai ht lines become curved



21/38

21


Resampling types for raster image

output raster image are different, interpolation is

required for each pixel.

Resampling methods:

Nearest NeighbourNearest Neighbour

Bilinear Interpolation

Cubic Interpolation

Cubic Interpolation


Green line original image


22/38

22



Nearest Neighbour


Pixel value(s) retained !


23/38

23



Bilinear Interpolation


Smooth result


24/38

24


Choice of resampling types

Spectral Spatial accuracy

Nearest neighbour : spectral information retained

u c n erpo a on : smoo er spa a n orma on



25/38

25


Images Georeferenced vs. Geocoded

Georeferenced ima e Geocoded ima e


Flevoland - IJssel delta, The Netherlands


26/38

26


Images Topomap vs . Geocoded image

To o ra hical ma Geocoded ima e


Flevoland - IJssel delta, The Netherlands


27/38

27


Three-dimensional approaches (1)

MonoplottingUsin a DEM to calculate terrain coordinates no re-sam lin

Ortho-image productionResampling of an image into map geometry taking terrainrelief into account

Stereo-restitutionUse of two images (stereo-pair) to extract 3D information

Orietation


Results in formula to calculate image coordinates (X, Y) fromterrain coordinates (x, y, z)


28/38

28



geospatial data describes the horizontal position of

terrain features : X, Y coordinates

displacement

3D geospatial data also considers also vertical positionof terrain features (for instance height above sea

level) : X, Y, Z coordinates



29/38

29



the ground surface

surface, including

vegetation, infra-


, .


30/38

30



recons ruc on o e pos on othe projection centre of sensor: AP-> Principal Point



31/38

31



: econs ruc on o pos onand attitude / inclination of the sensor with respect toterrain coordinate system.

Parameters:

1. Indirect camera orientation: Measurements of GCP

in the terrain X Y and Z

2. Direct camera orientation: Orientation of the

sensor itself (position and attitude) by GPS and


3. Integrated Camera orientation: Combi of 1. and 2.


32/38

32



:

RPC : Rational Polynomial Coefficient

coordinates of an entire frame and terraincoordinates.

Can be used by RS software, such as

ERDAS Imagine



33/38

33



onop o ng: s ng a o ca cu a e erra n

coordinates no resampling needed



34/38

34



r op o o: esamp ng o an mage n o map

geometry taking terrain relief into account

Monoplotting+


ResamplingOn the original aerial photo Geometrically correct

the map grid is distorted orthophoto


35/38

35



ereo res u on: se o wo mages s ereo-

pair) to extract 3D information. For measurements

use is made of arallax



36/38

36


Summary (1)

To correct mages 2 mens ona y or stort ons:

1. Georeferencing

2. Geocoding

Use of appropriate Transformation

Use of Ground Control Points (GCP) from GPS or

Topographical map

Transformation types:

Conformal: preservation of angles

Affine: reservation of arallels


Polynomical (2nd order)


37/38

37


Summary (2)

Image resampling methods:

Nearest Neighbour : pixel values retained

Bilinear Inter olation: smooth result

Cubic convolution Three dimensional approaches for correction of

Internal & external orientation

Monoplotting


Use of DEM


38/38

38


Questions


damen2010 geometric operations

Documents