Kwang-Hoon Chi

Geostatistical integration of spectral and spatial information for land-cover mapping using remote sensing data

A geostatistical contextual classifier for land-cover mapping using remote sensing data is presented. To integrate spatial information with spectral information derived from remote sensing data, a geostatistical indicator approach is adopted to determine the probability of a certain land-cover class occuring at an unsampled location given that any other land-cover classes occur at neighboring locations. The geostatistical indicator algorithm applied here is simple indicator kriging with local means. This approach can directly integrate both spatial information of ground data (hard data) and spectral information of remote sensing data (soft data) within an indicator kriging framework. This algorithm is applied to the classification of multi-sensor remote sensing data for land-cover mapping. This classification result is compared with a result obtained from a conventional spectral information based classification method.

Integration of IKONOS imagery for geostatistical mapping of sediment grain size at Baramarae beach, Korea

We describe the potential of high-resolution remote sensing imagery in the geostatistical mapping of sediment grain size distribution in order to supplement sparsely sampled ground observations. Within a multi-Gaussian framework, the IKONOS imagery is used as local means both to estimate the grain size values and to model local uncertainty at unsampled locations. Multiple regression and generalized additive models are applied to compute local mean values. From a case study of Baramarae beach, Korea, all imagery bands showed a reasonable linear relationship with grain size values in phi units, having a correlation coefficient of more than –0.80. Accounting for the IKONOS imagery via simple kriging with local means could reflect detailed surface characteristics with less smoothing effects. Cross validation results showed that the mean square errors from simple kriging with local means via the generalized additive model provided a relative improvement of about 60% over univariate multi-Gaussian kriging and a superior predictive capability when compared with simple kriging with local means via the traditional multiple regression model.

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.

Land-cover Classification using Multi-temporal/polarization C-band SAR Data

This paper presents a fuzzy logic fusion methodology for land-cover classification with multi-temporal/polarization Radarsat-1 and ENVISAT ASAR data. For feature extraction from each multi-temporal/polarization data, a traditional feature extraction approach (i.e. extraction of average backscattering coefficient, temporal variability and long-term coherence) and principal component analysis (PCA) were applied and compared. A data-driven fuzzy logic approach was applied to the classification of those features. In the fuzzy logic approach, fuzzy membership functions based on smoothed kernel density estimation and likelihood ratio functions were derived and various fuzzy combination operators were tested. A case study from an agricultural area has been carried out to illustrate the proposed methodology.

Application of fuzzy set theory for spatial prediction of landslide hazard

This work presents a fuzzy logic scheme to integrate multi-source spatial data for landslide hazard mapping. Main focus of This work is on fuzzy information representation based on frequency ratio and non-parametric density estimation. Of particular interest is the representation of continuous data for preventing loss of information. The non-parametric density estimation method applied here is a Parzen window estimation that can directly use continuous data without any categorization procedure. A case study at Jangheung, Korea is presented to illustrate the proposed scheme.