Change Detection Analysis of Urban Forest in Klang Valley using Multiemporal
Remote sensing Data: some Preliminary Result
Data Processing
-
Geometric Correction
Two multispectral Landsat Thematic Mapper senses acquired in 1991 and 1993 were both registered to a common database . to ensure high geometric accuracy is achieved, a RMSE of less than 0.5 pixel is tolerated in the geometric correction. The cubic Convolution was adopted in the resampling scheme. Digital image processing system EASI/PACE PCI was used in the data processing.
- Change Detection
change detection was carried out based on the analysis of the least-correlated components of the Principal Component Analysis ( PCA) output. In this technique, all raw bands ( totaling 14 bands ) were used as main input the PCA. A set of 14 correlated information's were transformed. Up to 95% of information were transformed into the first three highly -correlated components of the PCA output while the least-correlated in information were transformed to the last few components produced. The least-correlated components, in real world can reflect to changes that have taken place in the range of period where the data were captured.
Results
Fourteen components were generated in the PCA transformation. The eigenvalue, deviation and its variance percentage is given in Table 1.
Table 1 Eigenvalue and Variance of the 14 Generated Componrnts
| Eigenchannel |
Eigenvalue |
Deviation |
%variance |
| 1 |
4712.3721 |
68.6467 |
50.44% |
| 2 |
3397.9480 |
58.2919 |
36.37% |
| 3 |
769.8112 |
27.7455 |
8.24% |
| 4 |
160.3480 |
12.6629 |
1.72% |
| 5 |
93.3378 |
9.6611 |
1.00% |
| 6 |
70.8839 |
8.4193 |
0.76% |
| 7 |
56.3467 |
7.5064 |
0.60% |
| 8 |
23.0203 |
4.7980 |
0.25% |
| 9 |
17.1336 |
4.1393 |
0.18% |
| 10 |
12.5489 |
3.5424 |
0.13% |
| 11 |
11.8361 |
3.4404 |
0.13% |
| 12 |
9.9092 |
3.1479 |
0.11% |
| 13 |
5.7050 |
2.3885 |
0.06% |
| 14 |
1.8554 |
1.3621 |
0.02% |
