Applicability of Landsat ETM+ SLC-off Imagery Filled using the Neighborhood Similar Pixel Interpolator for Change Detection
Abstract
The scan-line corrector (SLC) of the Landsat 7 Enhanced Thematic Mapper Plus (ETM+) failed in 2003, which seriously limited scientific applications of ETM+ data. A method called the Neighborhood Similar Pixel Interpolator (NSPI) was developed to interpolate the values of the pixels within the gaps. This method is based on the assumption that the same-class neighboring pixels around the un-scanned pixels have similar spectral characteristics and that these neighboring and un-scanned pixels exhibit similar patterns of spectral differences between dates. Previous study indicates that NSPI performs well for filling image data gaps, even across heterogeneous landscapes. This study evaluated the applicability and effectiveness of using the NSPI-filled SLC-off ETM+ images for change detection. Six Landsat path/rows representing a variety of landscapes with different types of dominant disturbances were selected, and an SLC-on image pair was obtained for each path/row. A simulated SLC-off image was generated for one image from each pair and filled using NSPI. We employed the Multi-Index Integrated Change Analysis (MIICA) model to produce a change image with two classes-change and no change-from each image pair. Evaluation of the effectiveness of NSPI-filled images in change detection was made by comparing spectral change results derived from gap-filled simulated SLC-off image pairs with those from the original image pairs. In addition, one actual Landsat 7 SLC-off image and a Landsat 5 image acquired close in time to the Landsat 7 data set were chosen to further evaluate the applicability of NSPI-filled images for change detection. All gap-filled images were spatially and radiometrically continuous without obvious striping patterns. The change detection results from gap-filled simulated image pairs spatially agreed very well with those from original image pairs. Moreover, the spatial agreement between the change results derived from the gap-filled actual SLC-off image pair and an alternate "close-date" SLC-on image pair reaches 93% for all classes and 58% for the change class on a pixel-by-pixel basis within the entire scene, and 91% for all classes and 48% for the change class within the gap-filled pixel areas. These preliminary results indicate that there is great potential for applying NSPI-filled images for change detection to capture different disturbance types and will likely be useful for assessing fire, urban development, mining, and forest harvesting and regeneration.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2011
- Bibcode:
- 2011AGUFM.B51M0582J
- Keywords:
-
- 0480 BIOGEOSCIENCES / Remote sensing;
- 1640 GLOBAL CHANGE / Remote sensing