A Tutorial on Support Vector Machines for Pattern Recognition
Data Mining and Knowledge Discovery
Probability Estimates for Multi-class Classification by Pairwise Coupling
The Journal of Machine Learning Research
International Journal of Remote Sensing
| Hi-index | 0.00 |
The potential value of combining broadband and multispectral thermal infrared (TIR) data with multispectral and hyperspectral visible, near-infrared (VNIR) and shortwave infrared (SWIR) data was investigated within the context of urban land-cover classification. Using a case study of airborne Digital Airborne Imaging Spectrometer (DAIS) imagery of Strasbourg, France, the relative contribution of TIR wavelengths to classification accuracy was investigated for hyperspectral and simulated multispectral IKONOS, SPOT and Landsat Thematic Mapper (TM) bands. A support vector machines (SVM) classifier was used because this method was found to be very effective at handling the complex distributions of the heterogeneous land cover classes. The overall classification accuracy varied greatly with different band combinations. The inclusion of a single broad thermal band increased classification accuracy by as much as 20% for simulated IKONOS bands, but only 4% for hyperspectral VNIR and SWIR data. Adding multispectral TIR data raised the average accuracy approximately a further 10% for each band combination studied. Thermal wavelengths were found to be particularly useful for reducing the confusion between road and roof surfaces.