期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2016
卷号：III-3
页码：457-464
出版社：Copernicus Publications
摘要：An energetical approach is proposed for classification decision fusion in urban areas using multispectral and hyperspectral imageryat distinct spatial resolutions. Hyperspectral data provides a great ability to discriminate land-cover classes while multispectral data,usually at higher spatial resolution, makes possible a more accurate spatial delineation of the classes. Hence, the aim here is toachieve the most accurate classification maps by taking advantage of both data sources at the decision level: spectral properties of thehyperspectral data and the geometrical resolution of multispectral images. More specifically, the proposed method takes into accountprobability class membership maps in order to improve the classification fusion process. Such probability maps are available usingstandard classification techniques such as Random Forests or Support Vector Machines. Classification probability maps are integratedinto an energy framework where minimization of a given energy leads to better classification maps. The energy is minimized usinga graph-cut method called quadratic pseudo-boolean optimization (QPBO) with α-expansion. A first model is proposed that givessatisfactory results in terms of classification results and visual interpretation. This model is compared to a standard Potts modelsadapted to the considered problem. Finally, the model is enhanced by integrating the spatial contrast observed in the data source ofhigher spatial resolution (i.e., the multispectral image). Obtained results using the proposed energetical decision fusion process areshown on two urban multispectral/hyperspectral datasets. 2-3% improvement is noticed with respect to a Potts formulation and 3-8%compared to a single hyperspectral-based classification.
关键词：multispectral imagery; hyperspectral imagery; urban land cover; regularization; fusion; classification; energy minimization; graphical model
