rsproject final pres

24
02-Jun-14 1 1 1 MS - 04 (RS&GIS), Term Project COMPARISON OF PIXEL BASED & OBJECT BASED CLASSIFICATION TECHNIQUES FOR GLACIER CHANGE DETECTION 2

Upload: amir-javaid

Post on 11-Aug-2015

29 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: RSProject final pres

02-Jun-14

1

1

1

MS - 04 (RS&GIS), Term Project

COMPARISON OF PIXEL BASED & OBJECT BASED CLASSIFICATION TECHNIQUES FOR GLACIER CHANGE

DETECTION

2

Page 2: RSProject final pres

02-Jun-14

2

3

Group Members Amir Javaid

Sher Muhammad

Chaman Gul

Supervisor Mr. Mirza Muhammad Waqar (IST, Karachi)

Contents 4

Introduction

Literature Review

Purpose of the study

Objectives

Study Area

Data Required

Methodology

Results

Conclusion

Page 3: RSProject final pres

02-Jun-14

3

Introduction

Glaciers are important indicators of sustainable life on the globe by

various means and provide good motivation for continuous

monitoring (Bishop et al., 2000; Kargel et al., 2005).

Glaciers are vital and big resource of freshwater which is used by the

people in the form of agriculture and energy (Knight, 1999).

Glaciers accumulate in winters and ablate in summers. The normal

accumulation and melting phenomena of the glaciers is affected by the

changing climate which in turn seriously impact human life and

economy.

5

Mountain Glaciers

These glaciers develop in high mountainous regions, often flowingout of ice fields that span several peaks or even a mountain range.The largest mountain glaciers are found in Arctic Canada, Alaska,the Andes in South America, the Himalayas in Asia, and onAntarctica.

Valley Glaciers

Commonly originating from mountain glaciers or ice fields, theseglaciers spill down valleys, looking much like giant tongues. Valleyglaciers tend to be very long, often flowing down beyond the snowline, sometimes reaching sea level (study area glaciers are valleyglaciers).

6

Glacier Types

Page 4: RSProject final pres

02-Jun-14

4

7

Piedmont Glaciers

Piedmont glaciers occur when steep valley glaciers spill intorelatively flat plains, where they spread out into bulb-like lobes. TheMalaspina Glacier in Alaska, covering over 5,000 square kilometersis one of the most famous examples of this type of glacier.

Cirque Glaciers

Cirque Glaciers are named for the bowl-like hollows they occupy, which are called cirques. Typically, they are found high on mountainsides and tend to be wide rather than long.

Hanging Glaciers

Also called ice aprons, these glaciers cling to steep mountainsides. Like cirque glaciers, they are wider than they are long. Hanging glaciers are common in the Alps, where they often cause avalanches due to the steep inclines they occupy.

Land covers Classified

Snow

Precipitation in the form of solid, usually occurs when the temperatureis below 0 Centigrade.

Debris

As glaciers creep along the landscape, they often pick up pieces ofrock and transport them as the glacier advances. When the glaciermelts, these pieces of rock are left behind as glacial debris.

8

Page 5: RSProject final pres

02-Jun-14

5

Literature Review

In January 2012 research published “Pixel-based and

Object-oriented Classification of High Resolution Satellite Images”

by Y. Aruna Suhasini Devi and Dr.I.V.Murali Krishna, they concluded that

The results show that object-oriented classification can produce satisfying results when compared with pixel-based method.

The overall accuracy was 87.5% by object-oriented method, while that of pixel-based method gave an overall accuracy of 78.1%.

9

Literature Review

In 2011 work published on “COMPARISON OF PIXEL-BASED AND

OBJECT-ORIENTED CLASSIFICATION APPROACHES USINGLANDSAT-7 ETM SPECTRAL BANDS” by M. Oruc he Found that.

The result of the accuracy assessment shows that object orientedimage analysis attain higher overall accuracy and higher individualproducer’s and user’s accuracy for each classified land cover class.

In object base method smaller scale increases the dimensionality and dividing the object into the sub-groups, while the larger scale combines the multi segments into one.

10

Page 6: RSProject final pres

02-Jun-14

6

Literature Review

In 2007, paper published on “Study of Glaciers in Northern Pakistan” by M.Haq, R.Jilani they found

Area of Batura & Biafo glaciers decreased from 1992-2000

Decreasing trend of water in Indus river system.

In 2012 a study on “Climate Change Effect on Glacier Behavior” By Pandey and Venkataraman

Glacier will happily advancing in a healthy climate and retreating in response to a warmer climate.

11

12

In 2012, a research had been launched on, “Spatiotemporal distribution of snow in eastern Tibet and the response to climate change” by Jie Gao and found that

Duration of the snow-free period was inversely correlated withelevation" at lower elevation sites the length of the snow-freeseason increased, in contrast, at higher elevation, it decreased.

Analysis of local temperature and precipitation, snow-free periodchanges with elevation however median date of the snow freeperiod is quite stable.

Literature Review

Page 7: RSProject final pres

02-Jun-14

7

Literature Review 13

In 2009, worked on “Monitoring of Mountain Glacial Variations inNorthern Pakistan using Landsat and ALOS Data “ by M. Haq, R. Jilaniet. al found that

Temporal loss in snow cover area of Yazghil, Jutmau and PassuGlaciers.

Glaciated areas of the glaciers have decreased near their terminus.

Literature Review

Bayr et al., 1994; Gratton et al., 1990; Paul et al., 2002; Sidjak and

Wheate, 1999, found that

Remote Sensing experts have been developing a number of methods to

assess variation in area and volume of the glaciers.

Several methods such as supervised classification or thresholding of

ratio images are available to delineate the glacier ice.

14

Page 8: RSProject final pres

02-Jun-14

8

Literature Review

Related study by Benn and Evans, in 1998 stated that

The ice of glaciers is either clean or debris covered. The source of

debris on the glaciers may vary from place to place. Mainly the

activities which cause debris cover on glaciers are

a) mass movements from adjacent mountain slopes.

b) wind-blown dust.

c) pollutants.

15

Objectives16

Objective of the study was to detect temporal changes and findsuitable classification technique for clean and debris cover ice.

Monitoring of glaciers is difficult by direct field methods whereas,Remote sensing offers an efficient technique for glaciological studies.

Long term study in short period of time.

To analyze changes and map study area Glaciers.

Page 9: RSProject final pres

02-Jun-14

9

Objective of the study17

Glaciers variation is an important indicator of Climate change.

Observation of Glaciers retreat and advancement to support Hazardsmanagement.

For scientific investigation of climate change, the glaciers retreat andadvance study by using satellite data.

Study Area18

Page 10: RSProject final pres

02-Jun-14

10

19

Data and software Required20

Data Type/ Software Specification Source

LANDSAT satellite data dataset (1972-2011) http://earthexplorer.usgs.gov

http://glovis.usgs.gov

DEM Aster 30m Resolution http://earthexplorer.usgs.gov

ERDAS Imagine Image Processing Leica

Arc GIS Mapping ESRI

eCognition Image Processing Definien

Page 11: RSProject final pres

02-Jun-14

11

MethodologyFlow Chart

Data Acquisition

Raster: LANDSAT (MSS, TM) Imagery (1972 – 2011), ASTER DEM, Google Earth

Landsat Data

Layer Stacking

Geometric Correction

Atmospheric Correction

Google EarthDEM

Extraction of Glacier Boundary

Glacier Boundary Overlay Operation

Subset Image

Supervised Classification

Define Training Sits

Evaluate Signatures

Classification

Area Estimation

Knowledge Based Classification

Spectral Signatures

B4 NDSI

Making Decision Rules

Classification

Area Estimation

Multi -Resolution image segmentation

Define training sites

NDSI

Making Decision Rules

Classification

Area Estimation

Accuracy Assessment with Reference Classification

Object Oriented Classification

Output Statistics/ Charts/ Map Layouts

Results and Discussions22

Due to climate change glaciers vary according to its location andweather condition.

Multi-temporal analysis on glacier extent using satellite imagery is animportant and valuable tool to understand climate variability, sinceglaciers respond very fast to climate change.

Variation occurs near terminus of the glaciers

Valley Glaciers exposed to South retreats and to North are stable oradvances

Page 12: RSProject final pres

02-Jun-14

12

Results and Discussions

Different Classification techniques based on thresholds of opticalimages (LANDSAT, or Equivalent) is a valuable tool for multitemporalanalyses of glacier and snow-cover changes.

Debris-covered ice is difficult to map with different classificationswithout manual supervision, since it could be misclassified withrecently deglaciated terrain at high altitude.

Object based classification is best technique to study temporal changesof glaciers

23

24

Page 13: RSProject final pres

02-Jun-14

13

25

26

Page 14: RSProject final pres

02-Jun-14

14

27

28

Page 15: RSProject final pres

02-Jun-14

15

29

30

Page 16: RSProject final pres

02-Jun-14

16

31

32

Page 17: RSProject final pres

02-Jun-14

17

33

34

Page 18: RSProject final pres

02-Jun-14

18

35

36

Page 19: RSProject final pres

02-Jun-14

19

37

0

1

2

3

4

5

6

7

8

9

1972 1998 2010 1972 1998 2010 1972 1998 2010

Ice Debris Rocks

Are

a (s

q.k

m)

Change Detection of Sachen Glacier by Different Classification Techniques

Knowledge Based

Supervised

Object Based

Reference

38

0

1

2

3

4

5

6

7

8

9

1972 1998 2010 1972 1998 2010 1972 1998 2010

Ice Debris Rocks

Are

a (s

q.k

m)

Change Detection of Rupal Glacier by Different Classification Techniques

Knowledge Based

Supervised

Object Based

Reference

Page 20: RSProject final pres

02-Jun-14

20

39

0

10

20

30

40

50

60

70

80

1978 1998 2011 1978 1998 2011 1978 1998 2011

Ice Debris Rocks

Are

a (

sq.k

m)

Change Detection of Apsara Glacier by selected Classification Techniques

Knowledge Based

Supervised

Object Based

Reference

40

0

10

20

30

40

50

60

70

80

90

1978 1998 2011 1978 1998 2011 1978 1998 2011

Ice Debris Rocks

Are

a (S

q K

m)

Change Detection of Singhi Glacier by Different Classification Techniques

Knowledge Based

Supervised

Object Based

Reference

Page 21: RSProject final pres

02-Jun-14

21

Accuracy Assessment (%)41

Overall Accuracy

Results and conclusions

Throughout the study period, it was observed that, about 9% of area islost by Rupal and Sachen glaciers whereas, Apsara and Singhi have lost3% of its snow cover area.

Object based classification provide best results with comparison ofother classification techniques.

After object based, Supervised classification provide better results toclassify land-covers.

Knowledge base is not suitable for mixed pixel and to differentiatedebris and rocks.

80% to 90% results are matching and having more than 90% overallaccuracy in the study area.

42

Page 22: RSProject final pres

02-Jun-14

22

Recommendations

Field visit is required for accuracy assessment

Use and search for freely available software's to do object basedclassification

More strong technological software are required for image processing

Use object based techniques for land-cover classification rather thanother techniques.

Knowledge based and objects based is depended on rules so definethese rules carefully

Familiarization with image is necessary to differentiate land-covers forbetter results

43

Bibliography44

Bayr, Klaus J., D. K. Hall, and W. M. Kovalick. "Observations on glaciers in the eastern Austrian Alps using satellite data." International Journal of Remote Sensing, vol. 15, no. 9, pp. 1733-1742, 1994.

Bishop, M. P., Kargel, J. S., Kieffer, H., Machinnin, D. J., Raup., B. And Shroder, J. F., “Remote Sensing Science and Technology for Studying Glacier Processes in High Asia”, Annals Glaciology, vol. 31, pp. 164-170, 2000.

Gratton, D. J., Howarth, P. J. and Marceau, D.J., “Combining DEM parameters with Landsat MSS and TM imagery in a GIS for mountain glacier characterization”, IEEE Trans. Geosci. Remote Sens., vol. 28, pp. 766-769, 1990.

Haeberli, Wilfred, and Martin Hölzle, "Application of inventory data for estimating characteristics of and regional climate-change effects on mountain glaciers: a pilot study with the European Alps." Annals Glaciology, vol. 21, pp. 206-212, 1995.

Page 23: RSProject final pres

02-Jun-14

23

Bibliography

Paul, F., Huggel, C. and Kääb, A. “Combining satellite multispectral image data and a digital elevation model for mapping of debris-covered glaciers”, Remote Sensing Envr., vol. 89, pp. 510-518, 2004.

Sidjak, R. W. And Wheate, R. D., “Glacier mapping of the Illecillewaet Icefield, British Columbia, Canada, using Landsat TM and digital elevation data”, International Journal Remote Sens., vol. 20, pp. 273-284, 1999.

Jie Gao, “Spatiotemporal distribution of snow in eastern Tibet and the response to climate change”, 2012.

M. Haq, R. Jilani, “Study of Glaciers in Northern Pakistan”, 2007 .

Panday and Venkataraman, “Climate Change Effect on Glacier Behavior: A Case Study from the Himalayas”, 2012 .

M. Haq, R. Jilani, Monitoring of Mountain Glacial Variations in Northern Pakistan using Landsat and ALOS Data” 2009.

45

Benz, U. C., Hofmann, P., Willhauck, G., Lingenfelder, I. and Heynen, M.2003. “Multi-resolution, object-oriented fuzzy analysis of remotesensing data for GIS-ready information, ISPRS Journal ofPhotogrammetry & Remote Sensing”, 58 (2004) pp. 239-258v.

S. Shataee, T. Kellenberger, A.A. Darvishsefat, “Forest typesclassification using ETM+ data in the north of Iran/ Comparison ofObject-oriented with Pixel-based classification techniques”.

46

Bibliography

Page 24: RSProject final pres

02-Jun-14

24

47

48

Questions - Discussion