I use Landsat-8 image. I use image fusion techniques. And apply to pan band covers range of 0.50-0.68. And the same range covered by band B-2, G-3, R-4. And my study area covers portion of coastal (sea), river, and buildings etc., by using the above mentioned band can I go for general classification, (0r) what kind of feature can be classified by using the above band.
Clearly, relevant answers have been given. I would like just to add a bit more here. Theoretically, you can use bands 2, 3 and 4 (as BGR true color) of Landsat 8 images for classification. But the accuracy will not be ideal (high enough). Because the vegetation signal represented by the green band (band 3) is much lower than that by band 5 (NIR); and soil/rock information, if your study area is not purely covered with vegetation, will not be enough represented if you do not use the SWIR bands (bands 6 or 7). Furthermore, the solar radiation in SWIR regions are completely absorbed by water (if there is no atmospheric effects), and thus, using SWIR bands will give you purely classified water bodies (if you have them in your study area).
As Richard and Maria have both mentioned, it will be better to conduct the Pan-fusion to all MS bands, and using more bands, e.g., bands 5, 6 (or 7) for achieving your purpose. The accuracy will be higher and the result will be better as more land surface biophysical information has been taken into account.
If your computer is old (without much RAM and powerful CPU), you may try such band combinations: 752, 753, or 754, or 652, 653 and 654 to observe land surface from different aspects (soil, rock, water and vegetation). Please note that bands 234 (visible) are intercorrelated to each other (that is why I said that your result of classification would not be ideal), and so are the bands 6 and 7 (2 SWIR bands). Hence, there is no need to input all these bands that are intercorrelated for your classification. However, if your computer is powerful enough, you can input all these bands for your classification disregarding their intercorrelation.
In my opinion, Pan-fused MS bands can be used for land cover classification but it will not be reasonable to use them to produce vegetation indices such as NDVI, EVI for yield and biomass assessment, and change detection as the original values of pixels have been modified during fusion procedure.
I hope this may help you to clarifiy what you need to do in your study.
Yes you can use your data, and this wavebands generaly using for vegetation mapping, agricultural crop pattern mapping and chrolophyl content analyses.
in landsat 8 images, you have 2 blue bands, so the 432 color composite became like the 321 in landsat 7 as a natural light. you can use this for classification.
I guess bands 2, 3, 4 alone may be not enough for accurate classification. At least NIR band 5 should be included.
Clearly, relevant answers have been given. I would like just to add a bit more here. Theoretically, you can use bands 2, 3 and 4 (as BGR true color) of Landsat 8 images for classification. But the accuracy will not be ideal (high enough). Because the vegetation signal represented by the green band (band 3) is much lower than that by band 5 (NIR); and soil/rock information, if your study area is not purely covered with vegetation, will not be enough represented if you do not use the SWIR bands (bands 6 or 7). Furthermore, the solar radiation in SWIR regions are completely absorbed by water (if there is no atmospheric effects), and thus, using SWIR bands will give you purely classified water bodies (if you have them in your study area).
As Richard and Maria have both mentioned, it will be better to conduct the Pan-fusion to all MS bands, and using more bands, e.g., bands 5, 6 (or 7) for achieving your purpose. The accuracy will be higher and the result will be better as more land surface biophysical information has been taken into account.
If your computer is old (without much RAM and powerful CPU), you may try such band combinations: 752, 753, or 754, or 652, 653 and 654 to observe land surface from different aspects (soil, rock, water and vegetation). Please note that bands 234 (visible) are intercorrelated to each other (that is why I said that your result of classification would not be ideal), and so are the bands 6 and 7 (2 SWIR bands). Hence, there is no need to input all these bands that are intercorrelated for your classification. However, if your computer is powerful enough, you can input all these bands for your classification disregarding their intercorrelation.
In my opinion, Pan-fused MS bands can be used for land cover classification but it will not be reasonable to use them to produce vegetation indices such as NDVI, EVI for yield and biomass assessment, and change detection as the original values of pixels have been modified during fusion procedure.
I hope this may help you to clarifiy what you need to do in your study.
Once you target LULC of various matrices that respond differently to the electromagnetic spectrum, I would use all bands, but again and as Lars already mentioned, it depends on the classification algorithm you are using (parametric, non-parametric etc.). A variable selection method would also help in screening which band or combination of bands are more useful for your study case.
- Badges
- Science topic
- Similar topics
- Geoscience
- Geoinformatics