Pre-processing and classification of airborne hyperspectral data for wetlands mapping
The focus of this study is on information extraction from wetland areas using Pushbroom Hyperspectral Imager (PHI) data in Sungai Kisap, Langkawi. PHI data with high spatial and spectral resolution (2 m spatial and 5 nm spectral resolution) is very sensitive to variations in the reflectance of an ob...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 语言: | English |
| 出版: |
2004
|
| 主题: | |
| 在线阅读: | http://eprints.utm.my/id/eprint/4812/1/AlvinLauMengShinMFKSG2004.pdf |
| 标签: |
添加标签
没有标签, 成为第一个标记此记录!
|
| 总结: | The focus of this study is on information extraction from wetland areas using Pushbroom Hyperspectral Imager (PHI) data in Sungai Kisap, Langkawi. PHI data with high spatial and spectral resolution (2 m spatial and 5 nm spectral resolution) is very sensitive to variations in the reflectance of an object's surface. Therefore, hyperspectral data pre-processing (i.e. radiometric conection, geometric correction and data mosaicking, topographic normalization, data masking, spectral data reduction and spatial data reduction) were employed to ensure that data used for wetland information extraction are well corrected. To enable the feature extraction, two sets of spectral libraries (one for land cover classes and another for mangrove classes) were created fiom a field campaign. Feature d o n using a thresholding technique was employed to extract information fiom the PHI data. Two data classification techniques were also used, namely (1) Spectral Angle Mapper, and (2) Binary Encoding. Four land cover classes and four mangrove classes had been successfilly extracted Erom PHI data. A spectral Angle M' classified PHI image with spectral angle 0.3 radian gives the best classification result over 80 % of overall accuracy with Kappa Coefficient of 0.557. Other classifiers tested also give reasonable results (over 70% of overall accuracy). The final outputs of this study are a land cover map and a mangrove classes map of Sungai Kisap area. |
|---|