Comparison of five relative radiometric normalization techniques for remote sensing monitoring
Relative radiometric normalization (RRN) minimized radiometric differences among images caused by inconsistencies of acquisition conditions (such as weather, season, sensor, etc.) rather than change in surface reflectance. Five methods of RRN, i. e. image regression (IR), pseudo-invariant features (...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Zhejiang University Press
2005-05-01
|
| Series: | 浙江大学学报. 农业与生命科学版 |
| Subjects: | |
| Online Access: | https://www.academax.com/doi/10.3785/1008-9209.2005.03.0269 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Relative radiometric normalization (RRN) minimized radiometric differences among images caused by inconsistencies of acquisition conditions (such as weather, season, sensor, etc.) rather than change in surface reflectance. Five methods of RRN, i. e. image regression (IR), pseudo-invariant features (PIF), dark set-bright set normalization (DB), no-change set radiometric normalization (NC), and histogram matching (HM), were applied to 1993 and 2001 Landsat TM/ETM+image of Jiashan County for evaluating their performance in relation to land cover detection. No other parameters and variables but image pixel digital values were used, so these methods were very easy to apply, especially for historical remote sensing images. The root-mean-square error and the dynamic range were employed in comparing and evaluating the images normalized by five methods. A change detection algorithm, i. e., image subtraction, was applied to compare the effects on change detection. The results showed that DB worked best among the five methods at the study area, the PIF worked better. Finally, factors affecting the performance of relative radiometric normalization and the conditions of applying these methods were identified and discussed. |
|---|---|
| ISSN: | 1008-9209 2097-5155 |