Automatic Land Cover Classification for Regional Area by an Improved Supervised Learning Procedure

摘要

In the task of classification for remotely sensed imagery, it's always hard to get an automatic and accurate result. This paper proposed a supervised learning process for land cover classification. Feature learning and pattern learning are defined and combined for this classification process. First land cover features are extracted from many ancillary data, and object features are extracted from imagery. Then feature learning are adopted to select samples automatically. At last, a supervised classifier like decision trees is used to learn land cover pattern from these samples, and classify the imagery. In this process, all of the work are executed by machine. In the experiment, a SPOT5 imagery is used for land cover classification. The result shows that the proposed method classifies the remote sensing data with a high and stable accuracy.