Lecture 8 image pre-processing

Question 1

What does TOA mean

Answer 1

top of the atmosphere, means its uncorrected data

Question 2

What does BOA mean

Answer 2

bottom of the atmosphere therefore no influence of the atmosphere

Question 3

What are Aerosols - with examples

Answer 3

❑ ‘Aerosols’ – suspended solid particles
▪ dust, liquid water, particulate matter
▪ natural (e.g. fog, dust, wildfire) and anthropogenic (combustion processes) activity

Question 4

which wavelength are aerosols and molecules best at scattering and absorbing light efficiently.

Answer 4

❑ Both aerosols and molecules scatter light more efficiently at shorter wavelengths
▪ Blue wavelengths most affected
❑ Both aerosols and molecules absorb light
▪Depends on concentration of material

Question 5

what is atmospheric correction and why is it needed

Answer 5

gas molecules scatter and absorb light in the atmosphere, sometimes light from sun reflected back to sensor, sometimes its lost sometimes scattered then reaches sensor. atmospheric correction removes influences.
Atmospheric correction Process of removing the influence of the atmosphere on the imagery

Why? ▪ Quantitative analysis – relate reflectance to surface property
▪ Multispectral data for visual analysis
▪ Scattering increases inversely with wavelength
▪ Multi-temporal analysis – atmosphere changes with time

Question 6

If you are presented with a spectral signature of a forest what should it show?

Answer 6

Blue part of spectrum should have lowest value over vegetation. on this example blue is higher than green suggesting its not atmospherically corrected. Scattering effecting blue band most is rayle scattering - inversely proportional. blue waveband is most effected.

Question 7

An example of a histogram and soil

Answer 7

Amount of reflectance increases with wavelength, there is some scattering in the blue part of the spectrum due to the atmospheric influence.

Question 8

What are the methods of atmospheric correction

Answer 8

1. Dark object subtraction method
2. Empirical line method
3. Radiative transfer models

Question 9

Dark object subtraction

Answer 9

Identify pixel values of low reflectance areas (i.e. near zero) should have very low values:
❑ Dark shadows
❑ Clear water
If its not close to zero – the numbers must come from atmospheric scattering
* Lowest pixel values in visible and near-infrared are approximation to atmospheric path radiance
* Assumes surface should have zero reflectance
* ‘near zero’ value due to atm. scattering
* The values we are getting for Dark object are due to atmospheric path radiance
* Minimum values subtracted from image
Ideally band15 of MERIS should have zero reflectance for water
DN value 4 is due to atmospheric scattering, if this is subtracted from the image it will correct the effect of atmosphere

Question 10

what are the problems with Dark object subtraction

Answer 10

• Does not account for spatial variation in the atmosphere (full-scene correction). Ie doesn’t account for dust.
• Truly “dark” pixels are rare and dark pixels in all spectral bands are rarer. Still some reflectance everywhere.

Question 11

Describe the empirical line method

Answer 11

Selection of one dark and one bright target
* Ground reflectance measurement using field radiometer
* Sensor radiance computed from image is compared with ground reflectance
dark target = low reflectance, bright target = high reflectance – use to correct image.
Valid for a small area, cannot use for 100’s of km – normally used for correcting air craft data.

Question 12

Radiative transfer models

Answer 12

Radiative transfer models
* More complicated BUT more reliable
* Need more information to parameterize the model – Geometrical conditions (view/solar angles) – Atmospheric model for gaseous components (Rayleigh scattering)
* H2O, O3 , aerosol optical depth,  (transmittance / opacity)
– Aerosol model (type and concentration) (Mie scattering)
* Dust, soot, salt etc
. – Spectral condition
* bands and bandwidths
– Ground reflectance (type and spectral variation)
* surface scattering (default is to assume Lambertian….)
Require more input, this is what was used for assignment.

Question 13

Geometric Correction Geometric correction or georeferencing reasons why?

Answer 13

the process of transforming the x-y dimensions of a remotely sensed image so that it has the same scale and project properties of a selected map projection.
Ie, columns and rows in satellite to latitude and longitude – raster data
Reasons:
1. To transform an image to match a selected map projection
2. To be able to locate points and features of interest on the image
3. To be able to merge two or more remote sensing images to make a single image
4. To be able to merge two or more images of the same area collected at different times (images may have been collected by different sensors) 5. To be able to merge the remote sensing image with other geographic data with a similar projection (e.g., to use the satellite image as an information layer within a GIS

Question 14

Geometric Correction concept

Answer 14

*The map co-ordinate system is expressed in terms of northings and eastings with units of metres. *The image co-ordinate system is expressed in terms of rows and columns with units of pixels. *Unfortunately, the two co-ordinate systems do not coincide.

Question 15

Geometric Correction procedure

Answer 15

Ground Control Point (GCP): locations that are found both in the image and in the map
Matching points on an image with corresponding map (or image) coordinates (or GPS readings taken in field)
build up a matrix that relates coordinate system with coordinate system of the image – widely disperse GCP across image.
*The result of geo-correction is to produce a new output grid with its rows and columns aligned with the Eastings and Northings of a map projection system

Question 16

what is image resampling

Answer 16

Re-sampling is used to assign a DN value to the output matrix pixel – determined on the basis of the pixel values which surround its transformed position in the original input image matrix

Question 17

what are the types of image resampling

Answer 17

1.Nearest neighbour
2.Bilinear Interpolation
3.Cubic Convolution

Question 18

Describe nearest neighbour

Answer 18

1.Nearest neighbour Image pre-processing
*Uses the value of the closest input pixel for the output pixel value
*Retains the original value
* Important for quantitative analysis of pixel values
*Data values may be lost, while other values may be duplicate
Grids don’t align – moves to closest value. DN remains the same – doesn’t alter the value of the pixel.

Question 19

what is Bilinear Interpolation

Answer 19

2.Bilinear Interpolation
Uses the weighted average of the nearest four pixels to the output pixel
- dark grey areas = 4 grey areas original image, finds nearest pixel incorrected image and puts into output

More spatially accurate than nearest neighbour; output image is smoother, less blocky.
Changes the value, don’t always need pixel values, sometimes not much of a problem if need geometric - weather forecasting important to track trajectory of clouds, windspeed in atmosphere, shape is more important than value, correct by bilinear interpolaton. more accurate spatial data not just spectral variation.
But if you need quantitative data say biomass – this is a problem,

Question 20

What is cubic convolution

Answer 20

3.Cubic Convolution
Uses the weighted average of the nearest sixteen pixels to the output pixel
Can both sharpen image and smooth out noise
Takes longer to run, can improve the display of the image – good if interested in shape of image.

Question 21

Why apply geometric correction?

Answer 21

Overlaying satellite images with different types of spatial information – google earth
Geometric correction allows to present lots of other layers.

Question 22

Name 3 geometric correction resampling methods?

Answer 22

• Nearest neighbour – doesn’t change values
• Cubic convolution – changes values (averaging)
• Bilinear interpolation – changes values (gives average)
  Last two will maintain shape

Question 23

What two atmospheric constituents scatter and absorb EMR?

Answer 23

• Molecules – absorb energy (H20 and gases)
• Aerosols – scatter light (Dust, pollen, soot)

Question 24

What is dark object subtraction?

Answer 24

• Atmospheric correction method that assumes areas of low reflectance (e.g., dark shadows) should have zero reflectance (i.e., if they have a DN of 4 it is assumed that the reflectance (DN of 4) is the result of atmospheric scattering. )
• Issues – assumes atmosphere is uniform and doesn’t vary spatially, ok in very small areas not ok in larger areas.