|
Land Use
|
Land Cover Classification using CRL/NASDA PI-SAR Data
In order to produce a final land cover map appropriate for the ground truth data or actual
conditions, correctly classified classes from the three thematic layers derived from the
classification procedures (layer-1, layer-2 and layer-3) were integrated (sensitivity approach).
From layer-1 (unsupervised classification result), bare land, forest and bush land were accepted.
From layer-2 (supervised classification result), water, small structures and wetland were
accepted. From layer-3 (result of the textural image), urban area was accepted. The layers were
integrated with the overlay function of the ERDAS imagine 8.3. As a result, seven
representative land cover types were generated. Figure 6 and Table 3 shows resultant land cover
type distribution achieved from the sensitivity approach and its accuracy report, respectively.
The accuracy reports consisted of three measurements, producer's accuracy, user's accuracy, and
Kappa statistics. The results of the accuracy assessment indicated that the classification
procedure produced a relatively high percentage of correct results, with an overall Kappa
coefficient of 0.72 and overall classification accuracy of 76.7 per cent.
Fig. 6 Final land cover types derived from the PI-SAR image in the study area.
Table 3 Accuracy assessment report derived from sensitivity approach.
| Class Name |
Reference Totals |
Classified Totals |
Number of Correct |
Producers Accuracy(%) |
Users Accuracy(%) |
Kappa statistics |
| Water |
6 |
7 |
4 |
66.7 |
57.1 |
0.52 |
| Bare land |
5 |
7 |
5 |
100.0 |
71.4 |
0.68 |
| Wet land |
8 |
7 |
6 |
75.0 |
85.7 |
0.83 |
| Small structures |
6 |
7 |
5 |
83.3 |
71.4 |
0.68 |
| Bush land |
9 |
9 |
7 |
77.8 |
77.8 |
0.73 |
| Forest |
17 |
14 |
13 |
76.5 |
92.9 |
0.90 |
| Urban |
9 |
9 |
6 |
66.7 |
66.7 |
0.60 |
| Total |
60 |
60 |
46 |
- |
- |
|
| Overall accuracy |
76.7 |
0. 72 |
6. Conclusion
The relations found between land cover types and multi-polarization PI-SAR data show that it is
possible to use PI-SAR data to monitor land cover types. Compared with ground truth data and
available published data, an accuracy exceeding 76 per cent is possible if an adequate
classification procedure is used. This accuracy may be enough to produce land cover mapping
using multi-polarization images, including PI-SAR and ALOS PALSAR. Classification accuracy
may be improved if the multi-polarization data is combined with the multi-temporal data.
PI-SAR SAR data can be considered applicable to land cover type analysis. At present, PI-SAR
L-band is operating, and in the near future, a multi-polarization spaceborne SAR system, ALOS
PALSAR, will be launched. PALSAR will generate multi-polarization images in L-band. In this
study, the classification accuracy has demonstrated that land cover classification mapping was
possible using the L-band multi-polarization SAR data. According to Shimada (1999), global
forest mapping (including Southeast Asia, Central Africa, South America and Boreal Forest)
program will be carried out using PALSAR data, and detected deforestation will be compared
with the Global Mapping Program by JERS-1 SAR (GFMP) data. Therefore, the
multi-polarization SAR data have great potential in cloud covered tropical, sub-tropical and
humid areas because the backscatter level was in different for different polarizations, and the
backscatter level for specific land cover was also different for different polarizations. Based on
behavior of the backscattering coefficient for the different polarizations discussed above,
multi-polarization observation seems as effective for the land cover classification as optical
sensors.
References
- Aplin, P., Atkinson, P. M., and Curran, P. J., 1999. Fine Spatial Resolution Simulated Satellite
Sensor imagery for Land cover Mapping in the UK, Remote Sens. Environ., 68: 206-216.
- Congalton, R. G., 1991. A Review of Assessing the Accuracy of Classification of Remotely
Sensed Data, Remote Sens. Environ. 37: 35-46.
- ERDAS, ERDAS Field Guide, Fourth Edition, Revised and Expanded, ERDAS, Inc., Atlanta, Georgia, p. 35,
p. 152, 153.
- Qong, M. 1999. Evaluation of JERS-1 SAR Data for Vegetation Type in Arid regions. J. of
JSPRS, Vol.38, No. 4, pp. 4-16.
- Shimada, M., 1999. Development, calibration, validation and study plan of the higher product
processing algorithm of EORC. J. of JSPRS, Vol.38, No. 2, pp. 39-42.
- Tadono, T., Qong, M., Wakabayashi, H., Shimada, M., Kobayashi, T., and Shi, J., 1999.
Preliminary Studies for Estimating Surface Moisture and Roughness Based on a Simultaneous
Experiment with CRL/NASDA Airborne SAR (PI-SAR). 20 th , ACRS, Hong Kong.
- Wakabayashi, H., 1999. Potentiality, characteristics and possibility of PALSAR. J. of of JSPRS,
Vol.38, No. 2, pp. 31-38.
- Wakabayashi, H., Kobayashi, T., Satake, M., and Uratsuka, S., 1999. Airborne L-Band SAR
System: Charateristics and Initial Calibration Results. IGASS symposium, Hunburg, Germany.
|
|
|
|
|
