کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
4464769 | 1621827 | 2014 | 13 صفحه PDF | دانلود رایگان |
![عکس صفحه اول مقاله: Optimal attributes for the object based detection of giant reed in riparian habitats: A comparative study between Airborne High Spatial Resolution and WorldView-2 imagery Optimal attributes for the object based detection of giant reed in riparian habitats: A comparative study between Airborne High Spatial Resolution and WorldView-2 imagery](/preview/png/4464769.png)
• Geometric attributes are essential for the OBIA detection of giant reed.
• Giant reed has peculiar spectral and geometric traits which creates distinct objects.
• Optimal attributes for giant reed mapping depend on imagery characteristics.
• Airborne and satellite images have divergent profits for the detection of giant reed.
Giant reed is an aggressive invasive plant of riparian ecosystems in many sub-tropical and warm-temperate regions, including Mediterranean Europe. In this study we tested a set of geometric, spectral and textural attributes in an object based image analysis (OBIA) approach to map giant reed invasions in riparian habitats. Bagging Classification and Regression Tree were used to select the optimal attributes and to build the classification rules sets. Mapping accuracy was performed using landscape metrics and the Kappa coefficient to compare the topographical and geometric similarity between the giant reed patches obtained with the OBIA map and with a validation map derived from on-screen digitizing. The methodology was applied in two high spatial resolution images: an airborne multispectral imagery and the newly WorldView-2 imagery. A temporal coverage of the airborne multispectral images was radiometrically calibrated with the IR-Mad transformation and used to assess the influence of the phenological variability of the invader.We found that optimal attributes for giant reed OBIA detection are a combination of spectral, geometric and textural information, with different scoring selection depending on the spectral and spatial characteristics of the imagery. WorldView-2 showed higher mapping accuracy (Kappa coefficient of 77%) and spectral attributes, including the newly yellow band, were preferentially selected, although a tendency to overestimate the total invaded area, due to the low spatial resolution (2 m of pixel size vs. 50 cm) was observed. When airborne images were used, geometric attributes were primarily selected and a higher spatial detail of the invasive patches was obtained, due to the higher spatial resolution. However, in highly heterogeneous landscapes, the low spectral resolution of the airborne images (4 bands instead of the 8 of WorldView-2) reduces the capability to detect giant reed patches. Giant reed displays peculiar spectral and geometric traits, at leaf, canopy and stand level, which makes the OBIA approach a very suitable technique for management purposes.
Journal: International Journal of Applied Earth Observation and Geoinformation - Volume 32, October 2014, Pages 79–91