Collection and Interpretation of Remote Sensing Data, Kasper Johansen, University of Queensland

Collection and Interpretation of
Remote Sensing Data

600m|______| 140m|______| 70m|______|

Dr. Kasper Johansen, Email: k.johansen@uq.edu.au
Biophysical Remote Sensing Group
School of Geography, Planning and Environmental Management
The University of Queensland
GTAQ 2012 1

Objectives of talk

 To present how to collect and access
remote sensing image data and
introduce selected image interpretation
approaches and study exercises for
high school students

GTAQ 2012 2

Outline of talk

 Introduction to remote sensing
 Collection of remote sensing data
 Interpretation of remote sensing data:
 Short study on linking field and image data
 Student exercise 1
 Short study on image interpretation cues
 Student exercise 2
 Summary of this talk
 Questions and Resources
 Demonstration of Remote Sensing Toolkit for learning
purposes

GTAQ 2012 3

What is Remote Sensing and Why Use It

The science and art of obtaining information about an
object, area or phenomenon through the analysis of data
collected by a device that is not in contact with the
object, area or phenomenon under investigation
(Lillesand et al., 2004:1)
Not a Remote Sensing Measurement

GTAQ 2012 4


Rockhampton/Gladstone MODIS Image February 11, 2003
Source CSIRO

GTAQ 2012 5


The science and art of obtaining information about an
object, area or phenomenon through the analysis of data
collected by a device that is not in contact with the
object, area or phenomenon under investigation
(Lillesand et al., 2004:1)
Remote Sensing Measurement
Not a Remote Sensing Measurement

Rockhampton

Gladstone

Susp. sediment concentration

GTAQ 2012 6


GTAQ 2012 7

Applications: Cyclone Yasi

GTAQ 2012 8

Applications: Biomass mapping

by Satellite
Measured
Measured in field

GTAQ 2012 9

Application: Surface Temperature
http://earthobservatory.nasa.gov/IOTD/view.php?id=36699

GTAQ 2012 10

Application: Elevation mapping

Digital Elevation Model
 LIDAR

GTAQ 2012 11

 LiDAR (Light Detection and Ranging):
 LiDAR pulses from airborne transmitter
 Height difference between surface features = Time difference for returns
 High positional accuracy
 Very suitable for deriving vegetation structural and geomorphic
information

ENVM3201 - April 2011 12

 LiDAR data examples with high point density

ENVM3201 - April 2011 13


Predicted for 5.4 m

GTAQ 2012 14

Predicted flooding
Jan 2011

GTAQ 2012 15

Actual flooding
Jan 2011

GTAQ 2012 16

Application: Coral reef mapping

GTAQ 2012 17

GPS towed
Every dot is a photo on the reef by diver

Application: Coral reef mapping

GTAQ 2012 19

Remote Sensing Applications

 Response to increasing application areas =
increasing data dimensionality and availability
 Need to carefully select data and balance spatial
resolution, spectral resolution, temporal resolution,
acquisition costs and processing costs
 Always question where data comes from and how it
was derived - metadata
 Push towards public access to data sets and open
source processing tools to increase data sharing

GTAQ 2012 20

Remote Sensing Applications

GTAQ 2012 21

Collection of Remote Sensing Data

 How do you get access to remote sensing data
and what are the costs?
 High spatial resolution imagery:
 Geoimage, SKM, AAM, Fugro Spatial Solutions ~ $30/km2
 Airborne optical and LiDAR data:
 AAM, Fugro, AEROmetrex, ARA, Hyvista ~ $1500/km2
 Free Imagery:
 USGS EarthExplorer

GTAQ 2012 22


 Free Imagery:
 Google Earth – but not geo-referenced

GTAQ 2012 24


 Free Imagery:
 Google Earth – but not geo-referenced

 TERN Data Discovery Portal

GTAQ 2012 26

Interpretation of Remote Sensing
Data at Different Spatial Scales

GTAQ 2012 29

Outline of talk

GTAQ 2012 30

Outline of talk

GTAQ 2012 31

Outline of talk

GTAQ 2012 32

Case Study 1

Mapping Condition of Savanna Riparian
Zones in North Australia
GTAQ 2012 36

1. Tropical Savanna Riparian Zones
Australian tropical savannas Riparian zones

Source: Tropical Savannas CRC, 2003

GTAQ 2012 37

1. Importance of Riparian Zones

 Provision of stream shade

 Prevention of erosion

 Nutrient source from litter fall

 Natural filtering of pollutants

 Wildlife habitat

GTAQ 2012 38

2. Objective

To map biophysical parameters suitable
for assessing the environmental condition
of Australian savanna riparian zones at
local to regional scales based on the
integration of field survey and high spatial
resolution image data.

GTAQ 2012 39

3. Study Area – Daly River
2.4m pixels

Darwin

0.6m pixels

Approximate scale
Katherine 2km I____________I

Map of part of the Northern Territory QuickBird image of the Daly River study area

GTAQ 2012 40

3. Study Area – Daly River

GTAQ 2012 41

4. Methods - Field Survey Data
 Field measurements of 5m
x 5m quadrats on both
sides of transect line – 10m
wide transects
 Parameters:
1. Riparian zone width
2. River channel width
3. Percentage Canopy Cover
4. Leaf Area Index (LAI)
5. Ground cover
6. High impact weeds
7. Tree species
8. Bank stability
9. Flood damage
10. Vegetation overhang

GTAQ 2012 42


GTAQ 2012 43


GTAQ 2012 44


Approximate scale
300m I_______________I

GTAQ 2012 45

5. Results – Biophysical Models
Daly River 2005
1
y = 1.4419Ln(x) + 1.2253
0.9
R2 = 0.7893, n = 548
Percentage Canopy Cover

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
SAVI

GTAQ 2012 46

5. Results – Biophysical Maps
Pan-sharpened QuickBird Image
Percentage Canopy Cover Map

Approximate scale
100m I_________I

GTAQ 2012 47


GTAQ 2012 48

5. Results – Biophysical Models
Daly River 2005
6
y = 9.5382x - 4.4086
R2 = 0.7206, n = 548
5
Leaf Area Index

4

3

2

1

0
0 0.2 0.4 0.6 0.8 1
SAVI

GTAQ 2012 49

5. Results – Biophysical Maps
Leaf Area index Map
Pan-sharpened QuickBird Image

Approximate scale
100m I_________I

GTAQ 2012 50

4. Methods - Object-based classification
Processing sequence for object-based image classification

Original image Segmented image

Develop rule sets Classified image

GTAQ 2012 51

5. Results – Image Classification
Image Classification 2004
Multi-spectral QuickBird image Image Classification 2005
Multi-spectral QuickBird image
23 August 2004 – 13 August 2005
Classification Accuracy - 2004 Classification Accuracy - 2005
100 100
90 90
80 80
70 70
Percentage

Percentage
60 60
50 50
40 40
30 30
20 20
10 10
0 0
Cleared Water Savanna Riparian Transition Exposed Cleared Water Savanna Riparian Transition Exposed
areas zone zone banks areas zone zone banks
Land Cover Classes Land Cover Classes
Total number of samples = 350 Producer's Accuracy User's Accuracy Total number of samples = 350 Producer's Accuracy User's Accuracy

Approximate scale Approximate scale
500m I_________I 500m I_________I

GTAQ 2012 52

5. Results – Image Classification
River width of the Daly River - August 2005
100

90

80

River width (m)
70

60

50

40

30

20

10
Average river width = 46.51m
0
0 1082 2129 3446 4511 6582 8068 9239 10434 11548 12559 13767 14917 17267 19073
Distance (m)
Riparian zone width, west bank of the Daly River, 2005
140

120
Riparian zone width (m)

100

80

60

40

20
Average riparian zone width, west bank = 53.57m
0
0 962 1921 3270 5119 6548 8004 8907 9858 10888 11828 13723 15412 17247 18890
Distance (m)

GTAQ 2012 53

7. Object 2 - Results

GTAQ 2012 54

5. Results – Bank Stability Map
Pan-sharpenedDamage Map
Stream Bank Stability Image
Flood QuickBird Map

Approximate scale
Approximate scale
100m I_________I
100m I_________I

GTAQ 2012 55

6. Conclusions
 Indicators of riparian zone condition that can be
mapped with an accuracy feasible for multi-temporal
assessment:
 Percentage canopy cover
 Leaf area index
 Bank stability
 Flood damage
 Riparian zone width
 River width
 Large sample size of field data to improve
relationship between field and image based
measurements

GTAQ 2012 56

Study Exercise 1:
Considering Spatial Scale
 Aim
To understand how environmental features (e.g.
trees, buildings and landforms) are measured and
represented in remotely sensed images.

 Background
Any effective form of remote sensing requires in-
depth experience and measurement of the
environment you are working in. This suggested
field exercise with provide this link which will enable
a strong and realistic basis for image analysis and
interpretation skills.

GTAQ 2012 57

Study Exercise 1:

 Tasks
 Task 1: Position/Location using a GPS
 Task 2: What is in a pixel
 Task 3: Identifying features along a transect to match
up observations with image data
 Task 4: Comparing a high spatial resolution image
(e.g. Google Earth) with a Landsat image

GTAQ 2012 58

Study Exercise 1:
Task 1: Position/Location using a GPS
 Aim: To explain, demonstrate and measure
horizontal and vertical position using a global
positioning system (GPS) receiver
 Instrument: Hand held GPS receiver

 Basic Principles to Explain and Demonstrate:
 Measurements of horizontal and vertical position,
including map projections, coordinate systems, datum
 GPS principles
 Measurement accuracy

GTAQ 2012 59

Study Exercise 1:
 Record the GPS position of a single location or
feature at 1 minute intervals for 5 minutes
 Use the GPS receiver to accurately map the
boundary of two features at the field site

GTAQ 2012 60

Study Exercise 1:
Position Measurement Note taker:
EPE
Feature- Waypoint name
Name (estimated
Photo file Easting Northing Height
student: positional
name:
error)

Mean position
Standard deviation of
position

GTAQ 2012 61

Study Exercise 1:
Feature Mapping Note taker:
Feature 1 Feature- Photo file Easting Northing Height EPE Distance between
Waypoint name: Points
1)
2)
3)
4)
Feature 2 Feature- Photo file Easting Northing Height Distance between
Waypoint name: points
1)
2)
3)

i)

n)

GTAQ 2012 62

Study Exercise 1:

GTAQ 2012 63

Study Exercise 1:

http://www.earthpoint.us/ExcelToKml.aspx

GTAQ 2012 64

Study Exercise 1:
Task 2: What is in a pixel
 Aim: To measure and assess the effects of
increasing the pixel size of an imaging sensor
 Instrument: 2 x 50 m survey tapes, digital camara,
hand held GPS receiver, ranging poles
 Principles of multi-spectral optical imaging systems –
where do pixels come from
 What controls the size of features detectable in an image
 Level of spatial detail required for mapping
 Common imaging sensor pixel and scene dimensions

GTAQ 2012 65

Study Exercise 1:
 Use the two survey tapes to successively mark out
the boundaries of image pixels to be measured
 At each pixel size, take photos from the centre of
the pixel and record GPS corner coordinates
 For each pixel size record the number and
percentage coverage of different land cover types
(soil, concrete, grass, trees, etc.)

GTAQ 2012 66

Study Exercise 1:

GTAQ 2012 67

Study Exercise 1:
Note taker:
Pixel size - no: photo name Waypoint name Easting Northing Height
0.5 x 0.5 m 1
2
3
4
2.4 x 2.4m 1
5
6
7
10 x 10m 1
8
9
10
20 x 20 m 1
11
12
13
30 x 30 m 1
14
15
16
50 x 50 m 1
17
18
19 GTAQ 2012 68

Study Exercise 1:
Pixel composition: Note taker:
Pixel size - Surface Cover Type % of pixel Sketch (soil, concrete, grass, trees, asphalt,etc etc
covered
0.5 x 0.5 m

2.4 x 2.4m

10 x 10m

20 x 20 m

30 x 30 m

50 x 50 m

GTAQ 2012 69

Study Exercise 1:
Task 3: Identifying features along a transect to
match up observations with image data
 Aim: To measure and assess how land cover types
are represented in image data
 Instrument: 1 x 50 m survey tapes, digital camera,
hand held GPS receiver
 What does the satellite see
 What does a pixel look like when multiple land cover
types occur within it
 Why is there a need for integrating field and image data
(calibration and validation of image maps)
GTAQ 2012 70

Study Exercise 1:
Task 3:Identifying features along a transect to
match up observations with image data
 Locate a start point of the transect and lay out the
50 m tape
 Record the positions of the start and end points of
the transect using the GPS receiver
 Take photos along the transect

 Identify land cover types along the transect line and
make notes where along the transect these occur

GTAQ 2012 71

Study Exercise 1:

GTAQ 2012 72

Study Exercise 1:
Note taker:
Distance no: photo name Land cover Easting Northing Height
along transect type
0m - ? m 1
?m- ?m 2
3
4
5
6
7
8
9
10
11
12

 Display the location of the transect start and end points
in Google Earth
 Compare land cover observations with those identified
in Google Earth and explain any observed differences
GTAQ 2012 73

Study Exercise 1:
Task 4: Comparing a high spatial resolution image
with a Landsat image
 Aim: To compare two images with different spatial
resolutions
 Questions to Address:
 When and why would you use the two different image
types?
 What are the pros and cons of using the two different
image types?
 Think of different applications suitable for using the two
image types

GTAQ 2012 74

Study Exercise 1:

Landsat image Google Earth image
GTAQ 2012 75

Case Study 2

Object-Based Mapping of Urban Areas

GTAQ 2012 76

Urban Land Cover Mapping

 QuickBird image from 2005 500 250 0 500 Meters

GTAQ 2012 77


500 250 0 500 Meters

GTAQ 2012 78


500 250 0 500 Meters

GTAQ 2012 79


500 250 0 500 Meters

GTAQ 2012 80


500 250 0 500 Meters

GTAQ 2012 81


500 250 0 500 Meters

GTAQ 2012 82


500 250 0 500 Meters

GTAQ 2012 83


500 250 0 500 Meters

GTAQ 2012 84

Conclusions

 Object-based image analysis can be used to map
urban land cover classes at high spatial resolution

 Shape and size of objects and context relationships
were found very useful for mapping urban land cover
classes

GTAQ 2012 85

Study Exercise 2:
Interpreting images
 Aim
To build an understanding and experience in the
necessary skills for interpreting image data

 Background
Manual interpretation of aerial photos and high
spatial resolution image data is a well
established science. This science has recently
provided the basis for automated mapping
approaches using object-based image analysis

GTAQ 2012 86

Study Exercise 2:
Interpreting images
 Image Interpretation Cues
 Tone/Colour: bright / actual colour
 Texture: frequency of change and arrangement of
tones
 Size: physical size of objects
 Shape: shape created by the boundaries of features
 Shadows: presence and extent
 Pattern: repetition of shape and tonal features
 Context (site and association): geographic location
constraints of features (e.g. beaches near water),
positional association (e.g. aircraft, runway, airport)

GTAQ 2012 87

Study Exercise 2:
Interpreting images
 Image Interpretation Cues Terminology
 Tone/Colour: bright - dark / actual colour
 Texture: smooth - rough
 Size: physical size and dimensions of objects
 Shape: rectangular, circular, square, oval, etc.
 Shadows: presence and extent
 Pattern: regular - irregular
 Context (site and association): geographic location
constraints of features (e.g. beaches near water),
positional association (e.g. aircraft, runway, airport)

GTAQ 2012 88

Study Exercise 2:
Interpreting images
 Questions:
 Identify image interpretation cues for the following
land cover types: mangroves, canal estate, sugar
cane fields
 Identity which interpretation cues are unique for
certain land cover classes, which will allow
recognition and discrimination and different land
cover classes

GTAQ 2012 89

Study Exercise 2:
Interpreting images
Interpretation Land- Land- Land-
Cue cover/use #1 cover/use #2 cover/use #3
Tone –
Colour

Texture

Size

Shape

Pattern

Shadow

Context

GTAQ 2012 90

Study Exercise 2:
Interpreting images

GTAQ 2012 91

Study Exercise 2:
Interpreting images

GTAQ 2012 92

Study Exercise 2:
Interpreting images

GTAQ 2012 93

Summary of this talk
 Brief introduction about Remote Sensing

 Case study on relating field and image data

 Study exercise suitable for field trip

 Case study on automated use of image
interpretation cues

 Study exercise suitable for the classroom

 Further learning tools and resources
GTAQ 2012 94

www.gpem.uq.edu.au/cser-rstoolkit
GTAQ 2012 95

Collection and Interpretation of Remote Sensing Data, Kasper Johansen, University of Queensland

More Related Content

What's hot (20)

Similar to Collection and Interpretation of Remote Sensing Data, Kasper Johansen, University of Queensland (20)

More from becnicholas (15)

Recently uploaded (20)

Collection and Interpretation of Remote Sensing Data, Kasper Johansen, University of Queensland