Land Use/Cover Classification of Cloud-Contaminated Area by Multitemporal Remote Sensing Images

The increasing development of satellite remote sensing technology has provided a large amount of cheap and stable data sources for land cover/use observations. In mountainous area, it is usually to cloud-contained remote sensing images because of complex weather. Therefore, how to get land cover/use thematic maps in mountainous areas is a challenging topic. In this paper, an approach of classification for cloud-contained areas is proposed. The overall idea is described as follows. Firstly, investigate the variances between cloud cover areas and underlying surfaces, design classification methods with SVM, and implement precise detection of cloud cover areas. Secondly, use Kriging interpolation to build image inpainting models with time series landuse classification results. According to time series analysis theories, Kriging interpolation algorithm to enhance the precision in cloudcontained area will be built. Lastly, select a specific area and utilize domestic remote sensing images to test the feasibility and robustness of the proposed method and adjust model parameters.