Multi-sensor data fusion for supervised land-cover classification using Bayesian and geostatistical techniques

We propose a geostatistical approach incorporated to the Bayesian data fusion technique for supervised classification of multi-sensor remote sensing data. The classification based only on the traditional spectral approach cannot preserve the accurate spatial information and can result in unrealistic classification results. To obtain accurate spatial/contextual information, the indicator kriging that allows one to estimate the probability of occurrence of certain classes on the basis of surrounding pixel information is incorporated into the Bayesian framework. This new approach has its merit incorporating both the spectral information and spatial information and improves the confidence level in the final data fusion task. To illustrate the proposed scheme, supervised classification of multi-sensor test remote sensing data was carried out. Analysis of the results indicates that the proposed method considerably improves the classification accuracy, compared to the methods based on the spectral information alone.