lOMoARcPSD|34709425

Final Report Remote Sensing

Remote Sensing (Wageningen University & Research)

Studeersnel wordt niet gesponsord of ondersteund door een hogeschool of universiteit

Gedownload door Faissal Bozi ([email protected])
 lOMoARcPSD|34709425

MODULE 2

2-1 Give an example of a nominal or an ordinal scale.

A nominal scale is, for example the name of a city. An ordinal scale could be the
altitude of a point.

2-2 Give an example of an interval or a ratio scale.

An interval scale is, for example the temperature of an area (kelvin or degrees)
because we can understand the difference between the values.

2-3 Is it allowed to sum two land use maps of different years, where the land
use classes are numbered from 1 to 10? Why / Why not? Which file type should
be used to store these maps in Erdas Imagine?
No. The number is just a code to differ each land use, so this sum would have no
meaning. It is a nominal type.

2-4 If the values in an image have two discrete levels (two integer values),
which data type would you recommend for efficient storage? What is the
consequence of your choice in relation to the original data values?
We will choose the Signed 8 bit because the range of the data is between negative
and positive numbers and the bytes are not so big, so we do not expect many
unexpected results. The consequence of our choice is that we lose a big range of
data comparing to the other choices that we have in bigger bytes.

2-5 How would you store NDVI-values, which vary between 0 and 1?
What would the image look like if you store such an image as unsigned 8-bit?

Float Single would be more suitable in this situation. If it is stored in unsigned 8-bit it
would look like a 1-bit image.

2-6 Which data-format would you use if you mainly want to visualise single
bands of images?
We think that the best data format will be the BSQ, because the data are stored all
alongside each other for the same band in this format.

[1]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

2-7 What is the unit of file coordinates? Where is the origin of the coordinate
system in the image? What can be a unit of map coordinates? Do you always
have a point (0, 0) in your map coordinates within your image?
The unit for file coordinates are rows (x) and columns (y). The origin is top-left corner
of the image. For map coordinates can be either UTM or Lat/Long. In both situations
you don’t always have a 0,0 point.

2-8 A typical frame with remote sensing 16-bit data of the Landsat 8 OLI sensor
contains 5760 rows, 7020 columns and 8 spectral bands. Estimate the file size
in MB.
File size: 776,35584 MB

2-9 Assume that the values in 2 image data files are of the signed 8-bit type.
What is the possible range of output values if you divide (e.g. a/b) the
corresponding pixels of the 2 files? Which problem can come up?
Which data type would you recommend for the output file? Why?

The possible range would be an interval of length 256, because the division of -128
per -1 would be +128, which is not included in the 8-bit system. It would also include
decimal numbers. The problem was that these two mentioned above, making the
best output the single float, so it could represent better the result of the division.

2.1-1 Write down the range in pixel values present in e.g. a bright part of the
sky.

For the bright part of the sky the range is 127- 155 pixels.

2.2-1 Write down the section with Statistics Info. What is the Type of the file
and what is the Data Type?

Imagine image, unsigned 8 bit.

What is the meaning of the notions Mean, Median and Mode in the output?

Man is the average of the values, median is the value which divides the group of
values into 2 equal groups, Mode is the number which appears most times.

2.2-2 What do the values along the horizontal and vertical axis indicate?

The horizontal axis is DN value and vertical is the times that appears.

If you move the cursor inside the histogram area 3 values appear above the
histogram. What is the meaning of these values?
[2]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

How many times the DN values shows up.

2.2-3 Which grey tones in the image correspond with the lower DN-values and
which ones correspond with the higher DN-values ?

Dark grey corresponds to higher DN values.

2.2-4 Write down the minimum and maximum value.

MIN= 152, MAX= 157

Does it correspond with your answer of question 2.1-1 ?

Yes it does.

2.2-5 What happened with the pixel-data? What happened to the statistics?
Why does it have this effect on the appearance of the image?

The pixel data are now only in the range of the aoi layer that we selected, the
statistics are related to the aoi layer, because of the range of the DN values that it
has now.

2.4-1 Why do you need to specify the variables as float? What happens to the
output if you do not specify them correctly?

Everything will be unprecise if we don’t do them as float.

2.4-2 Which File Type and which Data Type for output will you choose? Why?

We select float single because is a division so we will have more accurate results, file
type: continuous because being a float there won’t be may equal results that can be
grouped in a thematic file.

2.4-3 Which land use types are emphasized?

The land use types which are emphasized is Vegetation.

2.4-4 Describe the data type of the input and output files. Copy the viewer with
the output image to your report, together with the function description window.

Both input and output are raster files.

[3]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

MODULE 3

[4]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

We used Landsat 8 because differently from the images of Landsat 7 it has

geometric pre-processing steps already performed on it, due to a problem Landsat 7
had in the final years of its activities.

We used the cloud cover filter 0% - 30%. In this range it is possible to do a decent
analysis of land cover and you still have plenty of images available. With less cloud
cover it is likely that not many images are available for a specific area.The time range
is between 01/10/2013 and 30/04/2014 for the first images, finding four scenes, and
from 01/06/2014 and 30/09/2014 for the comparison images, finding two scenes.
Row and path is defined as LC08_L1TP_197024_20140419_20170423_01_T1 for
the first period and LC08_L1TP_197024_20140606_20170422_01_T1 for the
second period.

Taking this information in consideration, we can assume that Landsat 8 is suitable for
the purpose of monitoring the land use cover, as it does not demand so high
resolution images neither frequent revisit time. The images being for free are also a
pro for this satellite.

For VHR images from SPOT 6 and 7 Satellite there were 373 images available for
the period from 2012 to 2013 and a cloud cover of until 30% even though most of the
images had no cloud cover. It has 1.5 m resolution in panchromatic sensor and for
other bands it can go until 10 m of spatial resolution, providing a very precise
classification of land use cover. According to the website it has revisit capacity daily
for any location due to its capability of capturing images from different angles, making
it suitable to provide the images necessary for monitoring the land cover with very
frequent updates. The storage demand would be a problem though, as these images
can be quite heavy. The price of each image would have to be taken in count too,
since it is probably not low.

STEP 2

Which metadata are available for the image

[5]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

Which errors can be visually identified in the image?

It is possible to see some clouds in the images and also errors due to extremely high
reflectance of greenhouse probably. It is also possible to note it is not geometrically
corrected comparing the image to a base map.

What is the current pre-processing level of the image?

The errors mentioned in the previous question evidence that, in a pre-processing
scale, it is a level 1 Image.

Which problems do you expect when the image is used for land use mapping?
For the classification of land use the lack of atmospheric correction would case
disruptions due the high reflectance in blue from the sky. The cloud cover would also
be a problem, turning the land cover harder to distinguish.

MODULE 4

4.1-1 Compare the image with the Topographic map Loenermark. Find out
which transformations should be applied to geocorrect the image. Describe
these transformations. What information do you need to apply this
transformation? Which steps should you subsequently make to find the
requested result?

[6]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

We use polynomial transformation. This transformation using the input coordinates

and reference coordinates in a polynomial equation
We need the coordinates of the GCP’s , then to use them in the polynomial equation.

4.1-2 How many ground control point (GCPs) do you need at least for an over
determined first order polynomial transformation?
We need at least 4 GCP’s.

Write down the estimated coordinates of the GCPs you read from the map in
meters (not kilometers).
We selected four points for our application. The first point is located in the
northwestern area of the map: X=.193.800. , Y=958.300.
The second point is located in the northeastern area of the map: X= 199300,
Y=457000.
The third point is located in the southeastern area of the map: X=197020, Y= 451240
The last point is located in the southwestern area: X= 193180 , Y= 453190.

What is the precision of the coordinates you derive from the map?
The precision is 20m.

4.1-3 Which co-ordinate-system are you going to ascribe to the image? rd new
What does the false easting and northing mean? False easting and northing
values are usually applied to ensure that all x and y values are positive. You can also
use the false easting and northing parameters to reduce the range of the x or y
coordinate values.
(http://help.arcgis.com/en/geodatabase/10.0/sdk/arcsde/concepts/geometry/coordref/
coordsys/projected/mapprojections.htm)

Where is the origin of this coordinate-system?

The apparent origin(0, 0) of the system lies 120 km southeast of Paris , on a field 1
km east of La Celle-Saint-Cyr(

What is the unit of the coordinate-system?

Meters

4.1-4 Write down the Control Point Errors of X, Y and Total.

Control point errors are for x= 39.9329, for y= 8.2611 and for the total= 40, 7785

In which units are the values expressed? Are the values tolerable ?
(Compare with the pixel size and the estimated precision in question 4.1-2)
The error is expressed in pixels. It is not tolerable

4.1-5 When is the nearest neighbor resampling technique recommended and

when the cubic convolution resampling technique?

[7]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

The nearest neighbor resampling is recommended when you are looking for extreme
values in the image and cubic convolution should be used for smoothing continuous
data. (https://support.esri.com/en/technical-article/000005606)

4.1-6 Remember that the default presentation of continuous image data is

determined by the mean and the standard deviation of the file statistics.
Explain why the option Ignore Zero in Stats. must be used.
It must be used to avoid the interference of errors in the image projection that could
show nonrealistic results in the statistics of pixel values later.

4.1-7 Check the coordinates of 1 easy to identify point (non-GCP) on the screen
with its coordinates you can read from the topographic map.
Write down the difference in X and Y direction you find.
The differences are 138 meters for the X and 8 meters to the Y

Are these values acceptable related to the Control Point Errors?

No because the differences are too high

4.2-1 How many ground control points do you need at least for an
overdetermined third degree polynomial transformation?
We need 11.

4.2-2 Write down for the 1st degree polynomial transformation, the Control
Point Errors of X, Y and Total. Give your comment on possible erroneous
GCPs.
For the 1st degree the error for x= 2.5159, y= 4.5558 and the total error= 5.2044. The
reason for that is lack of precision of the polynomial.

4.2-3 Write down for the 2nd degree polynomial transformation, the Control
Point Errors of X, Y and Total. Give your comment on possible erroneous
GCPs.
For the 2nd degree the error for x= 0.8581, y= 2.3811 and the total error= 2.5210.
This error is acceptable and the value for it is due to the imprecision in the caption of
the points.

4.2-4 Write down for the 3rd degree polynomial transformation, the Control
Point Errors of X, Y and Total. Give your comment on possible erroneous
GCPs. Which polynomial order for transformation is reasonable? Give an
explanation!
For the 3rd degree the error for x= 0.4914, y= 0.9051 and the total error= 1.0299.
The best one is the 3rd order because it gives better precision. The reason for that is
type of the equation.

[8]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

4.2-5 Can you trace spots or areas in the transformed image where the results
are not quite what you expected? Please give an explanation.
We run the inquire arrow throughout the image and apparently the results were quite
good in all of it.

MODULE 5
Make a graph of the spectral signature (based on corrected DN’s) for wheat for
the 3 moments of recording (plot the 3 curves in one figure).
· Make a comparable graph for potatoes.
· Make a comparable graph for town.
· Describe what you see in the graphs

The first image was captured in June, which means that the summer had just
begun. In this case crops are still being planted. It is notorious the higher value of
reflectance in August and September for all the bands, but mainly in band 4 (NIR).

[9]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

This phenomenon is caused by the increase of vegetation density as time advances

in summer. Even though one of the analysis is related to constructed areas, the
presence of arborization in the streets also can be sensed by band 4

Make a graph of the spectral signature for wheat for the 3 moments of
recording (plot the 3 curves in one figure). · Make a comparable graph for
potatoes. · Make a comparable graph for town.

[10]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

· What can you conclude from the graphs?

The conclusion is similar to the one in 5.1, the increase of vegetation density
is the reason for different values in the reflectance for the same band in different
months.

· What are the main differences with the graphs in task 5.1?
These graphs are probably more accurate because the correction applied on
the values is more complex. The differences in reflectance values are slightly lower,
wich means that the variation of atmospheric effects in the exact time of the images
capture was smoothened.

· Which object is most stable in time?

Constructed areas are the most stable due to less variability of land cover
along time

[11]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

MODULE 6

Which typical spectral signatures can you identify in the image (e.g., water,
soil, forest etc.)? Plot the signatures in a graph for your report.

WATER

FOREST

[12]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

SOIL

Evaluate both land surface and water. Especially for water focus on the lower
left part of the image (both river and lake). Plot the spectral profiles for this
area. What effects do you observe here and how can this be explained?

[13]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

In this parts of water bodies the angle of sun and airborne was the same, turning it in
a kind of mirror and reflecting directly to the sensor. The extremely high reflectance
saturates the sensor, which makes the value of some areas to be classified as ‘NO
VALUE’, returning black color in the display.

STEP 2

Choose a selection of bands from the visible, the NIR and the SWIR. What
effects do you observe for the different wavelength ranges?
In visible wavelength the hole image is darker except from some urban areas. Bare
soil is also brighter, but water and forest is dark.
In NIR, the image is lighter, mainly because of the forests that are brighter than
anything else. The water is still dark.
In SWIR, the image is in between of visible and NIR, with the vegetation not being
that dark but neither being as light as in NIR

Use the Image Metadata to assess the histograms for the individual bands.
Explain in the report the range of values which you find on the x-axis and the
y-axis. Go quickly through all the bands and describe the different histograms
types which are present in the dataset. Explain the histogram type for band 8,
band 17, band 25, band 40 and band 60.
The X axis is the DN values and Y axis is the number of pixels that has each value.
Band 8 is representing color red in visible spectrum. It is possible to notice that most
of the pixels have low reflectance values. This is explained by the spectral signature
of the land uses verified in the image, which have mostly low response in red.
Band 17 is in NIR spectrum. So that is why more pixels with high DN value show up
in the histogram, since vegetation has extremely high response on that part of the
spectrum. Soil has also important response on this wavelength.
In band 25 it is possible to notice most of pixels with low reflectance. The same effect
of NIR cannot be verified, since the spectral signature of vegetation drops in this part
of the spectrum.
The difference between 25, 40 and 60 is mostly due to the composition of the distinct
soils present in the image.

STEP 3

Make an estimation of the correlation coefficient between bands for the four
scatterplots, based on the shape of the point cloud in the scatterplot.
· Try to explain the difference in point clouds for the different scatterplots.
· Which errors in the image can you identify based on the scatterplots.
The correlation in the scatterplot of band 1 vs. band 4 is high, probably close to 1. as
bands 1 and 4 are in visible spectrum the reflectance of the land cover is quite
similar, providing this high correlation.
The correlation in the scatterplot of band 9 and 14 and 9 and 21 is moderate,
probability between 0 and 1. In the first case the difference of reflectance of
vegetation for red (band 9) and NIR (band 14) causes the distinct DN values
[14]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

frequency, resulting in the moderate correlation. It is possible to say the same thing
to the second scatterplot, even if it is a comparison of SWIR and red. The parts of it
that shows higher correlation is due to soil, that reacts in a more similar way in both
bands.
The correlation between bands 9 and 63 do not exist, being close to 0.
The errors in first three comparisons are probably motivated by the direct reflectance
of sun in the river and lake, being possible to identify them by points located in
unexpected areas of the scatterplot. I the last graph the cloud points’ shape evidence
an error in the band caption. Low energy waves are more likely affected by noise,
and band 63 is located at the last part of SWIR spectrum. The lack of correlation of
the scatterplot is explained by the low SNR of the image.

MODULE 7
Write down the used filter matrix.
The used filter matrix is 3*3.

7.2-2 Describe the visual differences between the original and the filtered
image. (Zoom in). Describe the results paying special attention to the changes
at edges and isolated peaks.
1. The image is smoothed.
2. In general the greater the matrix is, the effects in the image are less precise,
once they take in account a larger range of values

Write down the name of the filter and the filter matrix itself (Use the Edit
button). Describe the results of each filter paying special attention to the
changes at edges and isolated peaks.
3x3 Edge detect: the edges and the peaks are remarked
3x3 Low Pass: the edges are smoothed and the peaks of values are spreaded and
less bright
5x5 Horizontal: The horizontal edges are remarked and vertical ones are barely
visible. The peaks are spread in horizontal way
3x3 High Pass: the edges and the peaks are remarked
5x5 Summary: the edges and the peaks are remarked
5x5 Low Pass: the edges are smoothed and the peaks of values are spreaded and
less bright.

[15]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

Write down the name of this filter and explain how this filter works.
Describe the results of this filter paying special attention to the changes at
edges and isolated peaks.
Suppressing isolated noise or peaks; preserving edges. The median value of the 3x3
pixels matrix is applied do the central one.

Write down both filter matrices.

Which final operation is performed to create new pixel values?
The matrixes are 3x3.
The final operation is DIST, which computes distance from origin with SQRT (arg1**2
+ arg2**2).

Describe the results of the filter paying special attention to the changes at
edges and isolated peaks.
This filter is used for detection of edges and exaggeration of them. The peaks are
also remarked.

7.5-1 What is the file (not map !) row number?

The rows of the file are 335.

Show the output file in a Viewer. 7.5-3 Is the result satisfactory?

Yes the result is satisfactory as it seems below.

[16]

Gedownload door Faissal Bozi ([email protected])

 lOMoARcPSD|34709425

[17]

Gedownload door Faissal Bozi ([email protected])