Object-based image analysis through nonlinear scale-space filtering |
| |
| Authors: | Angelos Tzotsos Konstantinos Karantzalos Demetre Argialas |
| |
| Institution: | a National Technical University of Athens, Remote Sensing Laboratory, Athens, Greeceb Ecole Centrale de Paris, Laboratoire de Mathematiques Appliquees aux Systemes, Chatenay-Malabry, France |
| |
| Abstract: | In this research, an object-oriented image classification framework was developed which incorporates nonlinear scale-space filtering into the multi-scale segmentation and classification procedures. Morphological levelings, which possess a number of desired spatial and spectral properties, were associated with anisotropically diffused markers towards the construction of nonlinear scale spaces. Image objects were computed at various scales and were connected to a kernel-based learning machine for the classification of various earth-observation data from both active and passive remote sensing sensors. Unlike previous object-based image analysis approaches, the scale hierarchy is implicitly derived from scale-space representation properties. The developed approach does not require the tuning of any parameter—of those which control the multi-scale segmentation and object extraction procedure, like shape, color, texture, etc. The developed object-oriented image classification framework was applied on a number of remote sensing data from different airborne and spaceborne sensors including SAR images, high and very high resolution panchromatic and multispectral aerial and satellite datasets. The very promising experimental results along with the performed qualitative and quantitative evaluation demonstrate the potential of the proposed approach. |
| |
| Keywords: | Automation Analysis Simplification Segmentation Classification |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
