1- Shahrekord University , Ataollah.Ebrahimi@sku.ac.ir
2- Shahrekord University
2- Shahrekord University
Abstract: (93 Views)
Objective: During a landscape, it is not facile to discriminate land parts that have dissimilar amounts and types of vegetation. Plant Ecological Units (PEUs) are known as management units and are a reflection of the management actions and natural disturbances in the region. This research aims to fuse different resolutions of satellite images to increase the PEUs classification accuracy.
Methods: For this purpose, the Marjan-Borujen watershed in Chaharmahal va Bakhtiari province was selected. After field monitoring and surveys, four dominant PEUs groups were identified in the study area. In this study, bands from the Landsat_8 satellite images with 30 m spatial resolution (bands 7_2) and a 15 m panchromatic band (band 8) were used, as well as the Sentinel_2 satellite images including panchromatic bands (8, 4, 2, 3) with 10 m spatial resolution. First step, using the Landsat panchromatic band, the 30-m bands were upgraded to 15 m through the pen-sharpening process; so the 15 m data set was prepared from the Landsat_8 satellite. Then, to increase the spatial resolution of the 15-meter data set to 10 m, the Sentinel_2 panchromatic bands were used. In this way, the Sentinel_2 panchromatic bands were geometrically matched with the Landsat_8 15 m data set, and the Co-Registration process was performed with the minimum RMSE(0.05). Finally, two data sets (2 to 8 bands) of the Landsat_8 satellite images with 15 m and 10 m spatial resolution, the PEUs classification maps were prepared using the RF classification algorithm, and the maps' accuracy was displayed as an error matrix.
Results: The results show that increasing the spatial resolution significantly enhances the accuracy of PEUs classification maps. The 15 m set shows an overall classification map accuracy of 66%, while increasing the spatial resolution to 10 m enhances the overall accuracy to 82%. As well as, the error matrix results show that the classification map procured from the 10 m set, all four PEUs groups have improved the producer accuracy, user accuracy, and kappa agreement index. So, in this map, PEU 2 and PEU 3 have the highest kappa agreement coefficient (83 percent).
Conclusions: This study shows that using the Gram-Schmidt fusion algorithm and consequently increasing the spatial resolution of Landsat 8 images from 30 m to 10 m reduces mixed pixels and increases pure pixels, which in turn improves the quality of PEU classification maps.
Methods: For this purpose, the Marjan-Borujen watershed in Chaharmahal va Bakhtiari province was selected. After field monitoring and surveys, four dominant PEUs groups were identified in the study area. In this study, bands from the Landsat_8 satellite images with 30 m spatial resolution (bands 7_2) and a 15 m panchromatic band (band 8) were used, as well as the Sentinel_2 satellite images including panchromatic bands (8, 4, 2, 3) with 10 m spatial resolution. First step, using the Landsat panchromatic band, the 30-m bands were upgraded to 15 m through the pen-sharpening process; so the 15 m data set was prepared from the Landsat_8 satellite. Then, to increase the spatial resolution of the 15-meter data set to 10 m, the Sentinel_2 panchromatic bands were used. In this way, the Sentinel_2 panchromatic bands were geometrically matched with the Landsat_8 15 m data set, and the Co-Registration process was performed with the minimum RMSE(0.05). Finally, two data sets (2 to 8 bands) of the Landsat_8 satellite images with 15 m and 10 m spatial resolution, the PEUs classification maps were prepared using the RF classification algorithm, and the maps' accuracy was displayed as an error matrix.
Results: The results show that increasing the spatial resolution significantly enhances the accuracy of PEUs classification maps. The 15 m set shows an overall classification map accuracy of 66%, while increasing the spatial resolution to 10 m enhances the overall accuracy to 82%. As well as, the error matrix results show that the classification map procured from the 10 m set, all four PEUs groups have improved the producer accuracy, user accuracy, and kappa agreement index. So, in this map, PEU 2 and PEU 3 have the highest kappa agreement coefficient (83 percent).
Conclusions: This study shows that using the Gram-Schmidt fusion algorithm and consequently increasing the spatial resolution of Landsat 8 images from 30 m to 10 m reduces mixed pixels and increases pure pixels, which in turn improves the quality of PEU classification maps.
Send email to the article author
| Rights and permissions | |
 | This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
