1. Temporal Classification and Change Detection
March 20, 2000
Temporal Classification and Change Detection
FR 4262

2. 25. Temporal Classification and Change Detection
March 20, 2000
IKONOS Imagery
May 6
August 29
September 14
Rosemount Research & Outreach Center
FR 4262

3. Multitemporal Landsat 5 imagery
April
Multitemporal Landsat 5 imagery
May
June
July
Inter-temporal covariance provides separability not available in single date imagery
Numerous studies show improved classification accuracy with two or more dates of imagery over single date classifications

4. 25. Temporal Classification and Change Detection
March 20, 2000
February
Multitemporal IKONOS
Imagery
April
June
August
September
Cloquet Forestry Center
FR 4262

5. August, full-resolution image
25. Temporal Classification and Change Detection
August, full-resolution image
March 20, 2000
FR 4262

6. 25. Temporal Classification and Change Detection
March 20, 2000
May
FR 4262

7. 25. Temporal Classification and Change Detection
March 20, 2000
May
FR 4262

8. 25. Temporal Classification and Change Detection
March 20, 2000
September
FR 4262

9. 25. Temporal Classification and Change Detection
March 20, 2000
September
FR 4262

10. Use of Temporal Information in Classification
25. Temporal Classification and Change Detection
March 20, 2000
Use of Temporal Information in Classification
Basic rationale W A O
Time 1
Optimal time for W vs. A W A O
Time 2
Optimal time for W vs. O W A O
T1 + T2
All 3 classes are accurately classified
Inter-temporal covariance provides separability not available in single date imagery
FR 4262

11. Multitemporal Data and Classification Accuracy
25. Temporal Classification and Change Detection
March 20, 2000
Multitemporal Data and Classification Accuracy
Numerous studies with multitemporal data have shown that two or more dates are frequently better than a single date, especially if the single date is not at the optimal time
Classification Accuracy (%)
Acquisition Period
FR 4262

12. Monitoring Vegetation Dynamics with MODIS NDVI
25. Temporal Classification and Change Detection
March 20, 2000
Monitoring Vegetation Dynamics with MODIS NDVI
Movie of Minnesota "Green Up"
April – November 2006
FR 4262

13. 25. Temporal Classification and Change Detection
March 20, 2000
Temporal Profile Model (a second approach for using temporal information)
Several parameters can be derived from a profile model fit to several dates of data and used as features in image classification
Parameters are related to important biological-ecological characteristics
Reduces number of features to be classified
4 dates with 5 spectral bands (= 20 features) can be reduced to 4 or 5 features
Increases classification accuracy
FR 4262

14. Temporal Profile Parameters
25. Temporal Classification and Change Detection
March 20, 2000
Temporal Profile Parameters
start of green-up
rate of growth
maximum “greenness”
date of maximum
duration of greenness
rate of senescence
end of growing season
total seasonal accumulation (area under the curve)
Time
“Greenness” (NDVI) 3 4 6 2 5 8 1 7
FR 4262

15. Examples of Temporal Profiles for several Minnesota cover types
25. Temporal Classification and Change Detection
March 20, 2000
Examples of Temporal Profiles for several Minnesota cover types
FR 4262

16. 25. Temporal Classification and Change Detection
March 20, 2000
Example Images of Temporal Profile Metrics
Start Date (1)
Rate of Growth (2)
Maximum (3)
Time of Peak (4)
Duration (5)
Senescence
Rate (6)
AVHRR NDVI, 1998
FR 4262

17. 25. Temporal Classification and Change Detection
March 20, 2000
Garden City, Kansas
1972
1988
FR 4262

18. 25. Temporal Classification and Change Detection
March 20, 2000
1975
Rondonia, Brazil
1982
1992
FR 4262

19. 25. Temporal Classification and Change Detection
March 20, 2000
1975
Twin Cities Landsat Images
1981
1986
1991
1998
2002
Currently, change detection, monitoring and updating rely primarily on two techniques: image-to-image comparisons and post classification, or map-to-map, comparisons (Green et al, 1994). Image-to-image comparison methods apply various algorithms directly to multiple dates of satellite imagery (Ridd and Liu, 1998) to generate “change” vs. “no-change” maps. Map-to-map methods are considered “post classification” comparisons because they involve comparing two separate classifications of satellite data to produce “from-to” maps based on classification differences in the input maps (Yuan et al., 1998). Although subject to error propagation, map-to-map comparisons have several advantages over image-to-image comparisons. A multi-date series of land cover classifications can be used for many applications other than change detection, and the classifications enable deriving “from-to” change information.
FR 4262

20. General Steps for Digital Change Detection
25. Temporal Classification and Change Detection
March 20, 2000
General Steps for Digital Change Detection
Define the problem and select appropriate land cover classification system
Obtain appropriate imagery considering
spatial, spectral-radiometric and temporal resolution
atmospheric, illumination, seasonal, moisture, … conditions
Preprocess imagery
Geometric registration of multi-date images
Radiometric correction or normalization (depending on the classification approach)
Select and apply appropriate change detection algorithm
FR 4262

21. Classification of Image Differences
25. Temporal Classification and Change Detection
March 20, 2000
Date 1 imagery
Classification of Image Differences
Date 2 imagery
minus
Subtract one date of imagery from another to produce a “difference” image which is then classified
Difference images
“Change” map
Image classification
FR 4262

22. Classification of Image Differences
25. Temporal Classification and Change Detection
March 20, 2000
Classification of Image Differences
Advantages
Efficient way to detect change
Requires only one classification
Disadvantages
“From-to” change information is not available
Requires careful definition of “change - no change” threshold
differences in DN values due to other factors such as phenology, sun angle, atmosphere or sensors differences are not “real” changes
Requires acquisition of comparable imagery and careful radiometric calibration such that where there are no changes in land cover the images are near identical (i.e., difference equals zero)
FR 4262

23. Comparison of Classifications
25. Temporal Classification and Change Detection
March 20, 2000
Comparison of Classifications
Date 1 imagery
Classification of Date 1
Two dates are classified separately
Date 2 imagery
Classification of Date 2
FR 4262

24. Comparison of Classifications
25. Temporal Classification and Change Detection
March 20, 2000
Comparison of Classifications
Date 1 imagery
Classification of Date 1
Two dates are classified separately
Classification of Date 1
minus
Date 2 imagery
Classification of Date 2
Classification map of Date 2 is then subtracted from the map of Date 1
“Change” map
FR 4262

25. Comparison of Classifications
25. Temporal Classification and Change Detection
March 20, 2000
Comparison of Classifications
Advantages
provides “from - to” change class information
next base year is already completed
Disadvantages
accuracy of change map depends on the accuracy of the individual classifications
requires two classifications
FR 4262

26. 25. Temporal Classification and Change Detection
March 20, 2000
Summary
Multitemporal data, typically at different times of the year, can be used to increase classification accuracy and specificity
but does require acquiring and processing additional dates of data
Data acquired over different years can be used to detect and classify changes in land cover and use
Pre-classification vs. Post-classification change detection (from-to info).
FR 4262