Hyperspectral image classification by collaboration of spatial and spectral information

The classification of hyperspectral image data has drawn much attention in recent years. Consequently, it contains not only spectral information of objects, but also spatial arrangement of objects. The most established Hyperspectral classifiers are based on the observed spectral signal, and ignore the spatial relations among observations. Information captured in neighboring locations may provide useful supplementary knowledge for analysis. To combine the spectral and spatial information in the classification process, in this paper, a Multidimensional Local Spatial Autocorrelation (MLSA) is proposed for hyperspectral image data. Based on this measure, a collaborative classification method is proposed, which integrates the spectral and spatial autocorrelation during the decision-making process. The trials of our experiment are conducted on two scenes, one from HYDICE 210-band imagery collected over an area that contains a diverse range of terrain features and the other is toy car hyperspectral image captured at Instrumentation and Sensing Laboratory (ISL) at Beltsville Agricultural Research Center. Quantitative measures of local consistency (smoothness) and global labeling, along with class maps, demonstrate the benefits of applying this method for unsupervised and supervised classification.