论文部分内容阅读
Due to the improvement of remote sensing instruments, hyperspectral sensors can collect hundreds of channels simultaneously compares to multispectral imagery usually with less than ten channels. How to process this huge amount of data is a challenge problem, especially when the spectra of the land-covers are unavailable. In this study, we propose a fuzzy affinity propagation method for image classification. Affinity propagation is a clustering algorithm has fast execution speed and finds clusters with small error. Different distance measures can be used to estimate of how closely two pixel vectors resemble each other. After the affinity propagation clustering, a fuzzy classification is applied for soft classification based on the distance of every pixel vectors to each cluster. An AVIRIS image scene is adopted in experiment to demonstrate the capability of the proposed method.